| Overall Statistics |
|
Total Orders 220 Average Win 2.70% Average Loss -3.29% Compounding Annual Return 99.831% Drawdown 28.700% Expectancy 0.173 Start Equity 370000.00 End Equity 612746.53 Net Profit 65.607% Sharpe Ratio 1.766 Sortino Ratio 2.024 Probabilistic Sharpe Ratio 66.963% Loss Rate 36% Win Rate 64% Profit-Loss Ratio 0.82 Alpha 0.578 Beta -2.487 Annual Standard Deviation 0.389 Annual Variance 0.151 Information Ratio 1.81 Tracking Error 0.403 Treynor Ratio -0.276 Total Fees $0.00 Estimated Strategy Capacity $3800000.00 Lowest Capacity Asset USDJPY 8G Portfolio Turnover 378.32% |
Notebook too long to render.
#region imports
from AlgorithmImports import *
#endregion
## General Settings
general_setting = {
"tickers": {
"EURUSD": {"type": "forex"},
##-##
# "USDJPY": {"type": "forex"},
# "GBPUSD": {"type": "forex"},
# "AUDUSD": {"type": "forex"},
##-##
},
"model_name": "ForexLSTM_V1_0",
"consolidator_timeframes": ["D1", "W1"],
##-##
# "lstm_tickers": ['EURUSD','USDJPY','GBPUSD','AUDUSD'],
"lstm_tickers": ['EURUSD'],
##-##
"order_counter_diff": 3,
##-##
"signals": [
"FxLstm_Both_EURUSD",
],
# "signals": [
# "FxLstm_Both_EURUSD","FxLstm_Both_EURUSD_Trail",
# "FxLstm_Hybrid_EURUSD",
# "FxLstm_Both_USDJPY","FxLstm_Both_USDJPY_Trail",
# # "FxLstm_Hybrid_GBPUSD",
# "FxLstm_Hybrid_GBPUSD_Trail",
# "FxLstm_Both_AUDUSD",
# # "FxLstm_Both_AUDUSD_Trail",
# "FxLstm_Hybrid_AUDUSD", "FxLstm_Hybrid_AUDUSD_Trail",
# ],
##-##
"FxLstm_prediction_hour": 1,
"external_data": {},
# "external_data": {
# # SP500
# 'spy': {
# 'source': 'equity',
# 'ticker': 'SPY',
# },
# # Global X DAX Germany ETF
# 'dax': {
# 'source': 'equity',
# 'ticker': 'DAX',
# },
# # US Treasury
# 'us_treasury': {
# 'source': 'USTreasuryYieldCurveRate',
# 'ref': 'USTYCR',
# 'col_date': 'time',
# 'col_val': 'onemonth',
# },
# # Consumer Price Index for Inflation Rate
# # https://data.nasdaq.com/data/RATEINF-inflation-rates
# 'cpi_usa': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_USA",
# },
# 'cpi_eur': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_EUR",
# },
# 'cpi_deu': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_DEU",
# },
# # Federal Funds Effective Rate (DFF)
# # https://fred.stlouisfed.org/series/DFF
# 'dff': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vT5lyey5dhfrZifoZFuDwlQDOz6oILyUyAHTLVe2eqiLv9jWkNeIFITIeKqwBOtS8oEUOoZ2zXX1De7/pub?gid=1400614786&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'dff',
# 'lag_days': 2,
# },
# # Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for the Euro Area (19 Countries)
# # https://fred.stlouisfed.org/series/IRLTLT01EZM156N
# 'rate_eur_lt_gov': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSl_hxRnfcXnFly0Gh1vyZYNTRW6VTv-FQDlXuNUR1090RIst2a01nyhGl7tPR4VIcrgFfGBc3OSD72/pub?gid=1026565438&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IRLTLT01EZM156N'.lower(),
# 'lag_days': 10,
# },
# # Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for the Euro Area (19 Countries)
# # https://fred.stlouisfed.org/series/IR3TIB01EZM156N
# 'rate_eur_3m_bank': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSkVfnj8N9AIsVF5PJN0JzU9ahw71nK_sTwY2qLKtNNxs1JI0STexUPEW15dY9bDUN8Fwql7_WUiKhK/pub?gid=2059310805&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IR3TIB01EZM156N'.lower(),
# 'lag_days': 10,
# },
# # Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for Germany
# # https://fred.stlouisfed.org/series/IRLTLT01DEM156N
# 'rate_deu_lt_gov': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vToUOn242L-w9ZWUXz_fU59aUc6oN5tDJEG8fu207zO7jMyfy5y7VesxH0mzEKaqwuU7WGOq7_xxDSu/pub?gid=2099864712&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IRLTLT01DEM156N'.lower(),
# 'lag_days': 10,
# },
# # Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Germany
# # https://fred.stlouisfed.org/series/IR3TIB01DEM156N
# 'rate_deu_3m_bank': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTswIuhg3-tLwgP6RWSSPRyyLDpvHNqdlSgSNk91_SkUjKAD9_lyvhI84MAHRHzYdrIho1Narccx_w1/pub?gid=1568788544&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IR3TIB01DEM156N'.lower(),
# 'lag_days': 10,
# },
# },
"features": {
"D1": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10",
#"ULTOSC","CHOP","DX14","PHASE","CRSI","PSAR",
],
# "W1": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
# "D1": [],
"W1": [],
},
"features_val_map": {
"SMA10": ["val"],
"MACD": ["macd", "macdsignal", "macdhist"],
"ROC2": ["val"],
"MOM4": ["val"],
"RSI10": ["val"],
"BB20": ["upper","lower","mid"],
"CCI20": ["val"],
"ULTOSC": ["val"],
"CHOP": ["val"],
"DX14": ["val"],
"PHASE": ["val"],
"CRSI": ["val"],
"PSAR": ["val"],
},
}
## Consolidator Settings
consolidator_settings = {
"D1": {
"timeframe_minutes": 24 * 60,
"consolidation_type": "quote",
"indicators": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10","ATR14","ATR21",
"SMA5","SMA20","SMA50","SMA100","SMA200",
#"ULTOSC","CHOP","DX14","PHASE","CRSI","PSAR",
],
# "indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
"W1": {
"timeframe_minutes": 7 * 24 * 60,
"consolidation_type": "quote",
# "indicators": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
}
## Indicators Settings
indicator_settings = {
"SMA5": {
"type": "SMA",
"lookback": 5,
"field": "Close",
"window": 3,
},
"SMA10": {
"type": "SMA",
"lookback": 10,
"field": "Close",
"window": 3,
},
"SMA20": {
"type": "SMA",
"lookback": 20,
"field": "Close",
"window": 3,
},
"SMA50": {
"type": "SMA",
"lookback": 50,
"field": "Close",
"window": 3,
},
"SMA100": {
"type": "SMA",
"lookback": 100,
"field": "Close",
"window": 3,
},
"SMA200": {
"type": "SMA",
"lookback": 200,
"field": "Close",
"window": 3,
},
"MACD": {
"type": "MACD",
"window": 3,
},
"ROC2": {
"type": "ROC",
"lookback": 2,
"field": "Close",
"window": 3,
},
"MOM4": {
"type": "MOM",
"lookback": 2,
"field": "Close",
"window": 3,
},
"RSI10": {
"type": "RSI",
"lookback": 10,
"ma_type": "Simple",
"field": "Close",
"window": 3,
},
"BB20": {
"type": "BOLL",
"lookback": 20,
"ma_type": "Simple",
"std": 2,
"field": "Close",
"window": 3,
},
"CCI20": {
"type": "CCI",
"lookback": 20,
"field": "Close",
"window": 3,
},
"ATR10": {
"type": "ATR",
"lookback": 10,
"field": "Close",
"window": 3,
},
"ATR14": {
"type": "ATR",
"lookback": 14,
"field": "Close",
"window": 3,
},
"ATR21": {
"type": "ATR",
"lookback": 21,
"field": "Close",
"window": 3,
},
"ULTOSC": {
"type": "ULTOSC",
"window": 3,
},
"CHOP": {
"type": "CHOP",
"lookback": 52,
"window": 3,
},
"DX14": {
"type": "DX",
"lookback": 14,
"window": 3,
},
"PHASE": {
"type": "PHASE",
"lookback": 15,
"window": 3,
},
"CRSI": {
"type": "CRSI",
"rsi_len": 15,
"rsi_field": "Close",
"rsi_window": 21,
"window": 3,
},
"PSAR": {
"type": "PSAR",
"window": 3,
},
}
signal_settings = {
"FxLstm_Both_EURUSD": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
# "risk_pct": 0.01,
"risk_pct": 0.0125,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 21,
"longStopMultiplier": 0.1,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.1,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_EURUSD_Trail": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.01,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 21,
"longStopMultiplier": 0.1,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.1,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_EURUSD": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
# "risk_pct": 0.01,
"risk_pct": 0.0125,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 5.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_EURUSD_Trail": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.01,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 5.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_USDJPY": {
"lstm_ticker": "USDJPY",
"valid_tickers": ["USDJPY"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 1.0,
"shortStopMultiplier": 0.5,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 4.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_USDJPY_Trail": {
"lstm_ticker": "USDJPY",
"valid_tickers": ["USDJPY"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 1.0,
"shortStopMultiplier": 0.5,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 4.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_GBPUSD": {
"lstm_ticker": "GBPUSD",
"valid_tickers": ["GBPUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.40,
"shortStopMultiplier": 0.75,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_GBPUSD_Trail": {
"lstm_ticker": "GBPUSD",
"valid_tickers": ["GBPUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.40,
"shortStopMultiplier": 0.75,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_AUDUSD": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.75,
"shortStopMultiplier": 0.25,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.2,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_AUDUSD_Trail": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.75,
"shortStopMultiplier": 0.25,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.2,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_AUDUSD": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 1.0,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_AUDUSD_Trail": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 1.0,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
}
#region imports
from AlgorithmImports import *
#endregion
## General Settings
general_setting = {
"tickers": {
"EURUSD": {"type": "forex"},
},
"model_name": "ForexLSTM_V1_0",
"consolidator_timeframes": ["D1", "W1"],
"lstm_tickers": ['EURUSD'],
"order_counter_diff": 3,
# "signals": ["FxLstm_Both"],
# "signals": ["FxLstm_Hybrid"],
"signals": ["FxLstm_Both","FxLstm_Hybrid"],
"FxLstm_prediction_hour": 1,
"external_data": {},
# "external_data": {
# # SP500
# 'spy': {
# 'source': 'equity',
# 'ticker': 'SPY',
# },
# # Global X DAX Germany ETF
# 'dax': {
# 'source': 'equity',
# 'ticker': 'DAX',
# },
# # US Treasury
# 'us_treasury': {
# 'source': 'USTreasuryYieldCurveRate',
# 'ref': 'USTYCR',
# 'col_date': 'time',
# 'col_val': 'onemonth',
# },
# # Consumer Price Index for Inflation Rate
# # https://data.nasdaq.com/data/RATEINF-inflation-rates
# 'cpi_usa': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_USA",
# },
# 'cpi_eur': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_EUR",
# },
# 'cpi_deu': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_DEU",
# },
# # Federal Funds Effective Rate (DFF)
# # https://fred.stlouisfed.org/series/DFF
# 'dff': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vT5lyey5dhfrZifoZFuDwlQDOz6oILyUyAHTLVe2eqiLv9jWkNeIFITIeKqwBOtS8oEUOoZ2zXX1De7/pub?gid=1400614786&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'dff',
# 'lag_days': 2,
# },
# # Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for the Euro Area (19 Countries)
# # https://fred.stlouisfed.org/series/IRLTLT01EZM156N
# 'rate_eur_lt_gov': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSl_hxRnfcXnFly0Gh1vyZYNTRW6VTv-FQDlXuNUR1090RIst2a01nyhGl7tPR4VIcrgFfGBc3OSD72/pub?gid=1026565438&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IRLTLT01EZM156N'.lower(),
# 'lag_days': 10,
# },
# # Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for the Euro Area (19 Countries)
# # https://fred.stlouisfed.org/series/IR3TIB01EZM156N
# 'rate_eur_3m_bank': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSkVfnj8N9AIsVF5PJN0JzU9ahw71nK_sTwY2qLKtNNxs1JI0STexUPEW15dY9bDUN8Fwql7_WUiKhK/pub?gid=2059310805&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IR3TIB01EZM156N'.lower(),
# 'lag_days': 10,
# },
# # Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for Germany
# # https://fred.stlouisfed.org/series/IRLTLT01DEM156N
# 'rate_deu_lt_gov': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vToUOn242L-w9ZWUXz_fU59aUc6oN5tDJEG8fu207zO7jMyfy5y7VesxH0mzEKaqwuU7WGOq7_xxDSu/pub?gid=2099864712&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IRLTLT01DEM156N'.lower(),
# 'lag_days': 10,
# },
# # Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Germany
# # https://fred.stlouisfed.org/series/IR3TIB01DEM156N
# 'rate_deu_3m_bank': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTswIuhg3-tLwgP6RWSSPRyyLDpvHNqdlSgSNk91_SkUjKAD9_lyvhI84MAHRHzYdrIho1Narccx_w1/pub?gid=1568788544&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IR3TIB01DEM156N'.lower(),
# 'lag_days': 10,
# },
# },
"features": {
"D1": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10",
#"ULTOSC","CHOP","DX14","PHASE","CRSI","PSAR",
],
# "W1": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
# "D1": [],
"W1": [],
},
"features_val_map": {
"SMA10": ["val"],
"MACD": ["macd", "macdsignal", "macdhist"],
"ROC2": ["val"],
"MOM4": ["val"],
"RSI10": ["val"],
"BB20": ["upper","lower","mid"],
"CCI20": ["val"],
"ULTOSC": ["val"],
"CHOP": ["val"],
"DX14": ["val"],
"PHASE": ["val"],
"CRSI": ["val"],
"PSAR": ["val"],
},
}
## Consolidator Settings
consolidator_settings = {
"D1": {
"timeframe_minutes": 24 * 60,
"consolidation_type": "quote",
"indicators": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10","ATR14","ATR21",
"SMA5","SMA20","SMA50","SMA100","SMA200",
#"ULTOSC","CHOP","DX14","PHASE","CRSI","PSAR",
],
# "indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
"W1": {
"timeframe_minutes": 7 * 24 * 60,
"consolidation_type": "quote",
# "indicators": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
}
## Indicators Settings
indicator_settings = {
"SMA5": {
"type": "SMA",
"lookback": 5,
"field": "Close",
"window": 3,
},
"SMA10": {
"type": "SMA",
"lookback": 10,
"field": "Close",
"window": 3,
},
"SMA20": {
"type": "SMA",
"lookback": 20,
"field": "Close",
"window": 3,
},
"SMA50": {
"type": "SMA",
"lookback": 50,
"field": "Close",
"window": 3,
},
"SMA100": {
"type": "SMA",
"lookback": 100,
"field": "Close",
"window": 3,
},
"SMA200": {
"type": "SMA",
"lookback": 200,
"field": "Close",
"window": 3,
},
"MACD": {
"type": "MACD",
"window": 3,
},
"ROC2": {
"type": "ROC",
"lookback": 2,
"field": "Close",
"window": 3,
},
"MOM4": {
"type": "MOM",
"lookback": 2,
"field": "Close",
"window": 3,
},
"RSI10": {
"type": "RSI",
"lookback": 10,
"ma_type": "Simple",
"field": "Close",
"window": 3,
},
"BB20": {
"type": "BOLL",
"lookback": 20,
"ma_type": "Simple",
"std": 2,
"field": "Close",
"window": 3,
},
"CCI20": {
"type": "CCI",
"lookback": 20,
"field": "Close",
"window": 3,
},
"ATR10": {
"type": "ATR",
"lookback": 10,
"field": "Close",
"window": 3,
},
"ATR14": {
"type": "ATR",
"lookback": 14,
"field": "Close",
"window": 3,
},
"ATR21": {
"type": "ATR",
"lookback": 21,
"field": "Close",
"window": 3,
},
"ULTOSC": {
"type": "ULTOSC",
"window": 3,
},
"CHOP": {
"type": "CHOP",
"lookback": 52,
"window": 3,
},
"DX14": {
"type": "DX",
"lookback": 14,
"window": 3,
},
"PHASE": {
"type": "PHASE",
"lookback": 15,
"window": 3,
},
"CRSI": {
"type": "CRSI",
"rsi_len": 15,
"rsi_field": "Close",
"rsi_window": 21,
"window": 3,
},
"PSAR": {
"type": "PSAR",
"window": 3,
},
}
signal_settings = {
"FxLstm_Both": {
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.01,
"enter_long_trades": True,
"enter_short_trades": True,
# "sma_filter_lookback_fast": 5,
# "sma_filter_lookback_slow": 20,
"sma_filter_lookback": 100,
"atrLength": 21,
"longStopMultiplier": 0.1,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 1,
"movement_thres": 0.002,
"use_movement_thres_for_stops": False,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid": {
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.01,
"enter_long_trades": True,
"enter_short_trades": True,
# "sma_filter_lookback_fast": 5,
# "sma_filter_lookback_slow": 20,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 1,
"movement_thres": 0.002,
"use_movement_thres_for_stops": False,
"use_prediction_direction_to_exit": False,
},
}
#region imports
from AlgorithmImports import *
#endregion
## General Settings
general_setting = {
"tickers": {
"EURUSD": {"type": "forex"},
"USDJPY": {"type": "forex"},
"GBPUSD": {"type": "forex"},
"AUDUSD": {"type": "forex"},
},
"model_name": "ForexLSTM_V1_05",
"consolidator_timeframes": ["D1", "W1"],
"order_counter_diff": 3,
"model_types": ["both","hybrid"],
"lstm_tickers": ['EURUSD','USDJPY','GBPUSD','AUDUSD'],
"lstm_model_training_displace_days": {
'EURUSD': 0,
'AUDUSD': 0,
'USDJPY': 1,
'GBPUSD': 1,
},
"signals": [
"FxLstm_Both_EURUSD","FxLstm_Both_EURUSD_Trail",
"FxLstm_Hybrid_EURUSD",
"FxLstm_Both_USDJPY","FxLstm_Both_USDJPY_Trail",
# "FxLstm_Hybrid_GBPUSD",
"FxLstm_Hybrid_GBPUSD_Trail",
"FxLstm_Both_AUDUSD",
# "FxLstm_Both_AUDUSD_Trail",
"FxLstm_Hybrid_AUDUSD", "FxLstm_Hybrid_AUDUSD_Trail",
],
"FxLstm_prediction_hour": 1,
"external_data": {
# SP500
'spy': {
'source': 'equity',
'ticker': 'SPY',
},
# Global X DAX Germany ETF
'dax': {
'source': 'equity',
'ticker': 'DAX',
},
# US Treasury
'us_treasury': {
'source': 'USTreasuryYieldCurveRate',
'ref': 'USTYCR',
'col_date': 'time',
'col_val': 'onemonth',
},
# Consumer Price Index for Inflation Rate
# https://data.nasdaq.com/data/RATEINF-inflation-rates
'cpi_usa': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_USA",
},
'cpi_eur': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_EUR",
},
'cpi_deu': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_DEU",
},
'cpi_gbr': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_GBR",
},
'cpi_chf': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_CHE",
},
'cpi_jpn': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_JPN",
},
'cpi_can': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_CAN",
},
'cpi_aus': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_AUS",
},
# Federal Funds Effective Rate (DFF)
# https://fred.stlouisfed.org/series/DFF
'dff': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vT5lyey5dhfrZifoZFuDwlQDOz6oILyUyAHTLVe2eqiLv9jWkNeIFITIeKqwBOtS8oEUOoZ2zXX1De7/pub?gid=1400614786&single=true&output=csv",
'col_date': 'date',
'col_val': 'dff',
'lag_days': 1,
},
# Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/IRLTLT01EZM156N
'rate_eur_lt_gov': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSl_hxRnfcXnFly0Gh1vyZYNTRW6VTv-FQDlXuNUR1090RIst2a01nyhGl7tPR4VIcrgFfGBc3OSD72/pub?gid=1026565438&single=true&output=csv",
'col_date': 'date',
'col_val': 'IRLTLT01EZM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/IR3TIB01EZM156N
'rate_eur_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSkVfnj8N9AIsVF5PJN0JzU9ahw71nK_sTwY2qLKtNNxs1JI0STexUPEW15dY9bDUN8Fwql7_WUiKhK/pub?gid=2059310805&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01EZM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for Germany
# https://fred.stlouisfed.org/series/IRLTLT01DEM156N
'rate_deu_lt_gov': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vToUOn242L-w9ZWUXz_fU59aUc6oN5tDJEG8fu207zO7jMyfy5y7VesxH0mzEKaqwuU7WGOq7_xxDSu/pub?gid=2099864712&single=true&output=csv",
'col_date': 'date',
'col_val': 'IRLTLT01DEM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Germany
# https://fred.stlouisfed.org/series/IR3TIB01DEM156N
'rate_deu_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTswIuhg3-tLwgP6RWSSPRyyLDpvHNqdlSgSNk91_SkUjKAD9_lyvhI84MAHRHzYdrIho1Narccx_w1/pub?gid=1568788544&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01DEM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Switzerland
# https://fred.stlouisfed.org/series/IR3TIB01CHM156N
'rate_chf_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vRRVvpohXIZOGQ4HpAjTTMeZ6cTat0wZ1gOxpUR_5E3pDuDCHDppiRnV9GQNK33jWJ3pYxAjvOvmerO/pub?gid=1734297228&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01CHM156N'.lower(),
'lag_days': 10,
},
# 3-Month or 90-day Rates and Yields: Interbank Rates for Japan
# https://fred.stlouisfed.org/series/IR3TIB01JPM156N
'rate_jpn_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTjcVbe63Ea3BoVxTpTBNcaEICdI11DhVmZ6Qxb-_GcuP8VbemKreHWNEu5id0ZviHPk7PAtLHqdBGr/pub?gid=1682849610&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01JPM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Australia
# https://fred.stlouisfed.org/series/IR3TIB01AUM156N
'rate_aus_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vS73ca8pMDndu3lH5SjmrIJS-HwWfDdqS2mh1YQkQwhGj3UtIauP12xjhmLusXag9ibZJE3YZsWLERT/pub?gid=1639615970&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01AUM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Canada
# https://fred.stlouisfed.org/series/IR3TIB01CAM156N
'rate_cnd_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vQ0Odb11l33qCVwPS6G2lxrUpfQ5DWXnGw8HFu6uUV_OUx7b-yBIQItN12TwLRTq3Bx3-fBe-pU86ve/pub?gid=483123093&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01CAM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for United Kingdom
# https://fred.stlouisfed.org/series/IR3TIB01GBM156N
'rate_gbp_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTtQSIEiSK1swVM_oElodv6YsjzojdxwfXZm-hDx68DQD6V3HjtuOYpHb4KUQC5uWDFoe9t09-Mibex/pub?gid=1120780951&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01GBM156N'.lower(),
'lag_days': 10,
},
},
"features": {
"D1": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20","PSAR",
],
"W1": [],
},
"features_val_map": {
"SMA10": ["val"],
"MACD": ["macd", "macdsignal", "macdhist"],
"ROC2": ["val"],
"MOM4": ["val"],
"RSI10": ["val"],
"BB20": ["upper","lower","mid"],
"CCI20": ["val"],
"ULTOSC": ["val"],
"CHOP": ["val"],
"DX14": ["val"],
"PHASE": ["val"],
"CRSI": ["val"],
"PSAR": ["val"],
},
}
## Consolidator Settings
consolidator_settings = {
"D1": {
"timeframe_minutes": 24 * 60,
"consolidation_type": "quote",
"indicators": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10","ATR14","ATR21",
"PSAR",
"SMA100",
"SMA200",
],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
"W1": {
"timeframe_minutes": 7 * 24 * 60,
"consolidation_type": "quote",
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
}
## Indicators Settings
indicator_settings = {
"SMA5": {
"type": "SMA",
"lookback": 5,
"field": "Close",
"window": 3,
},
"SMA10": {
"type": "SMA",
"lookback": 10,
"field": "Close",
"window": 3,
},
"SMA20": {
"type": "SMA",
"lookback": 20,
"field": "Close",
"window": 3,
},
"SMA50": {
"type": "SMA",
"lookback": 50,
"field": "Close",
"window": 3,
},
"SMA100": {
"type": "SMA",
"lookback": 100,
"field": "Close",
"window": 3,
},
"SMA200": {
"type": "SMA",
"lookback": 200,
"field": "Close",
"window": 3,
},
"MACD": {
"type": "MACD",
"window": 3,
},
"ROC2": {
"type": "ROC",
"lookback": 2,
"field": "Close",
"window": 3,
},
"MOM4": {
"type": "MOM",
"lookback": 2,
"field": "Close",
"window": 3,
},
"RSI10": {
"type": "RSI",
"lookback": 10,
"ma_type": "Simple",
"field": "Close",
"window": 3,
},
"BB20": {
"type": "BOLL",
"lookback": 20,
"ma_type": "Simple",
"std": 2,
"field": "Close",
"window": 3,
},
"CCI20": {
"type": "CCI",
"lookback": 20,
"field": "Close",
"window": 3,
},
"ATR10": {
"type": "ATR",
"lookback": 10,
"field": "Close",
"window": 3,
},
"ATR14": {
"type": "ATR",
"lookback": 14,
"field": "Close",
"window": 3,
},
"ATR21": {
"type": "ATR",
"lookback": 21,
"field": "Close",
"window": 3,
},
"ULTOSC": {
"type": "ULTOSC",
"window": 3,
},
"CHOP": {
"type": "CHOP",
"lookback": 52,
"window": 3,
},
"DX14": {
"type": "DX",
"lookback": 14,
"window": 3,
},
"PHASE": {
"type": "PHASE",
"lookback": 15,
"window": 3,
},
"CRSI": {
"type": "CRSI",
"rsi_len": 15,
"rsi_field": "Close",
"rsi_window": 21,
"window": 3,
},
"PSAR": {
"type": "PSAR",
"window": 3,
},
}
signal_settings = {
"FxLstm_Both_EURUSD": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
# "risk_pct": 0.01,
"risk_pct": 0.0125,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 21,
"longStopMultiplier": 0.1,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.1,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_EURUSD_Trail": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.01,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 21,
"longStopMultiplier": 0.1,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.1,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_EURUSD": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
# "risk_pct": 0.01,
"risk_pct": 0.0125,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 5.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_EURUSD_Trail": {
"lstm_ticker": "EURUSD",
"valid_tickers": ["EURUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.01,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 0.2,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 5.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_USDJPY": {
"lstm_ticker": "USDJPY",
"valid_tickers": ["USDJPY"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 1.0,
"shortStopMultiplier": 0.5,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 4.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_USDJPY_Trail": {
"lstm_ticker": "USDJPY",
"valid_tickers": ["USDJPY"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 1.0,
"shortStopMultiplier": 0.5,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 4.0,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_GBPUSD": {
"lstm_ticker": "GBPUSD",
"valid_tickers": ["GBPUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.40,
"shortStopMultiplier": 0.75,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_GBPUSD_Trail": {
"lstm_ticker": "GBPUSD",
"valid_tickers": ["GBPUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 100,
"atrLength": 10,
"longStopMultiplier": 0.40,
"shortStopMultiplier": 0.75,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_AUDUSD": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.75,
"shortStopMultiplier": 0.25,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.2,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Both_AUDUSD_Trail": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'both',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.75,
"shortStopMultiplier": 0.25,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.2,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_AUDUSD": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
# "risk_pct": 0.005,
"risk_pct": 0.0075,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 1.0,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
"FxLstm_Hybrid_AUDUSD_Trail": {
"lstm_ticker": "AUDUSD",
"valid_tickers": ["AUDUSD"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.005,
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 200,
"atrLength": 21,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 1.0,
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": True,
"trailStopSize": 0.5,
"use_movement_thres_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
}
model_settings = {
"col_date": ['datetime'],
"col_price": 'close_D1',
"col_price_cur": 'price',
"col_target": 'target',
"start_year": 2013,
"trade_hour": 1,
"prediction_lookforward_days": 1,
"max_window_size": 100,
"scaled_tickers": ["USDJPY"],
"inflation_map_dict": {
'inflation_usa': 'cpi_usa',
'inflation_eur': 'cpi_eur',
'inflation_deu': 'cpi_deu',
'inflation_gbr': 'cpi_gbr',
'inflation_chf': 'cpi_chf',
'inflation_jpn': 'cpi_jpn',
'inflation_can': 'cpi_can',
'inflation_aus': 'cpi_aus',
},
"cols_data": [
'datetime', 'close_D1', 'price',
'spy', 'dax',
'dff',
'cpi_usa', 'cpi_eur', 'cpi_deu', 'cpi_gbr','cpi_chf','cpi_jpn','cpi_can','cpi_aus',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
'rate_chf_3m_bank',
'rate_jpn_3m_bank',
'rate_aus_3m_bank',
'rate_cnd_3m_bank',
'rate_gbp_3m_bank',
'D1-SMA10-val', 'D1-MACD-macd', 'D1-MACD-macdsignal', 'D1-MACD-macdhist', 'D1-ROC2-val',
'D1-MOM4-val', 'D1-RSI10-val', 'D1-BB20-upper', 'D1-BB20-mid', 'D1-BB20-lower', 'D1-CCI20-val',
'D1-PSAR-val',
],
'col_fundamental': [
'spy','dax',
'dff',
'cpi_usa','cpi_eur','cpi_deu','cpi_gbr','cpi_chf','cpi_jpn','cpi_can','cpi_aus',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
'rate_chf_3m_bank',
'rate_jpn_3m_bank',
'rate_aus_3m_bank',
'rate_cnd_3m_bank',
'rate_gbp_3m_bank',
],
'col_technical': [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
'D1-PSAR-val',
],
"model_settings_both": {
"use_gru_model": True,
"use_dual_lstm": False,
"epochs": 1,
"hidden_size": 50,
"window_size": 5,
"thres_multiplier": 3,
"use_early_stop": False,
"learning_rate": 0.0005,
"batch_size": 8,
"use_weighted_sampler": False,
"volatility_type": 'thres_auto_v1',
"valid_lookback_months": 12,
"train_lookback_months": 48,
"col_feature_dict": {
"EURUSD": [
'dff',
'inflation_usa_d30',
'inflation_eur_d30',
'inflation_deu_d30',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
'D1-PSAR-val',
],
"USDJPY": [
'dff',
'inflation_usa_d30',
'inflation_jpn_d30',
'rate_jpn_3m_bank',
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
'D1-PSAR-val',
],
"GBPUSD": [
'dff',
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
],
"AUDUSD": [
'dff',
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
'D1-PSAR-val',
],
},
},
"model_settings_hybrid": {
"use_gru_model": True,
"use_dual_lstm": False,
"epochs": 1,
"hidden_size": 50,
"window_size": 20,
"thres_multiplier": 3,
"learning_rate": 0.001,
"batch_size": 8,
"use_weighted_sampler": False,
"volatility_type": 'thres_auto_v1',
"valid_lookback_months": 12,
"train_lookback_months": 48,
"col_feature_fundamental_dict": {
"EURUSD": [
'dff',
'inflation_usa_d1',
'inflation_eur_d1',
'inflation_deu_d1',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
],
"USDJPY": [
'dff',
'rate_jpn_3m_bank',
],
"GBPUSD": [
'dff',
'inflation_usa_d1',
'inflation_gbr_d1',
'rate_gbp_3m_bank',
],
"AUDUSD": [
'dff',
'inflation_usa_d1',
'inflation_aus_d1',
'rate_aus_3m_bank',
],
},
"col_feature_technical_dict": {
"EURUSD": [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
],
"USDJPY": [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
],
"GBPUSD": [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
],
"AUDUSD": [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
],
},
},
}
#region imports
from AlgorithmImports import *
#endregion
## General Settings
general_setting = {
"tickers": {
##-##
"EURUSD": {"type": "forex"},
# "USDJPY": {"type": "forex"},
# "USDCHF": {"type": "forex"},
# "GBPUSD": {"type": "forex"},
# "AUDUSD": {"type": "forex"},
# "USDCAD": {"type": "forex"},
##-##
},
"model_name": "ForexLSTM_V1_0",
"consolidator_timeframes": ["D1", "W1"],
##-##
"lstm_tickers": ['EURUSD'],
# "lstm_tickers": ['USDJPY'],
# "lstm_tickers": ['USDCHF'],
# "lstm_tickers": ['GBPUSD'],
# "lstm_tickers": ['AUDUSD'],
# "lstm_tickers": ['USDCAD'],
##-##
"fx_lstm_signal_name": 'FxLstm',
"order_counter_diff": 3,
"external_data": {
# SP500
'spy': {
'source': 'equity',
'ticker': 'SPY',
},
# Global X DAX Germany ETF
'dax': {
'source': 'equity',
'ticker': 'DAX',
},
# US Treasury
'us_treasury': {
'source': 'USTreasuryYieldCurveRate',
'ref': 'USTYCR',
'col_date': 'time',
'col_val': 'onemonth',
},
# Consumer Price Index for Inflation Rate
# https://data.nasdaq.com/data/RATEINF-inflation-rates
'cpi_usa': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_USA",
},
'cpi_eur': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_EUR",
},
'cpi_deu': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_DEU",
},
'cpi_gbr': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_GBR",
},
'cpi_chf': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_CHE",
},
'cpi_jpn': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_JPN",
},
'cpi_can': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_CAN",
},
'cpi_aus': {
'source': 'NasdaqDataLink',
'ref': "RATEINF/CPI_AUS",
},
# Federal Funds Effective Rate (DFF)
# https://fred.stlouisfed.org/series/DFF
'dff': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vT5lyey5dhfrZifoZFuDwlQDOz6oILyUyAHTLVe2eqiLv9jWkNeIFITIeKqwBOtS8oEUOoZ2zXX1De7/pub?gid=1400614786&single=true&output=csv",
'col_date': 'date',
'col_val': 'dff',
'lag_days': 2,
},
# Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/IRLTLT01EZM156N
'rate_eur_lt_gov': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSl_hxRnfcXnFly0Gh1vyZYNTRW6VTv-FQDlXuNUR1090RIst2a01nyhGl7tPR4VIcrgFfGBc3OSD72/pub?gid=1026565438&single=true&output=csv",
'col_date': 'date',
'col_val': 'IRLTLT01EZM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/IR3TIB01EZM156N
'rate_eur_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSkVfnj8N9AIsVF5PJN0JzU9ahw71nK_sTwY2qLKtNNxs1JI0STexUPEW15dY9bDUN8Fwql7_WUiKhK/pub?gid=2059310805&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01EZM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for Germany
# https://fred.stlouisfed.org/series/IRLTLT01DEM156N
'rate_deu_lt_gov': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vToUOn242L-w9ZWUXz_fU59aUc6oN5tDJEG8fu207zO7jMyfy5y7VesxH0mzEKaqwuU7WGOq7_xxDSu/pub?gid=2099864712&single=true&output=csv",
'col_date': 'date',
'col_val': 'IRLTLT01DEM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Germany
# https://fred.stlouisfed.org/series/IR3TIB01DEM156N
'rate_deu_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTswIuhg3-tLwgP6RWSSPRyyLDpvHNqdlSgSNk91_SkUjKAD9_lyvhI84MAHRHzYdrIho1Narccx_w1/pub?gid=1568788544&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01DEM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Switzerland
# https://fred.stlouisfed.org/series/IR3TIB01CHM156N
'rate_chf_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vRRVvpohXIZOGQ4HpAjTTMeZ6cTat0wZ1gOxpUR_5E3pDuDCHDppiRnV9GQNK33jWJ3pYxAjvOvmerO/pub?gid=1734297228&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01CHM156N'.lower(),
'lag_days': 10,
},
# 3-Month or 90-day Rates and Yields: Interbank Rates for Japan
# https://fred.stlouisfed.org/series/IR3TIB01JPM156N
'rate_jpn_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTjcVbe63Ea3BoVxTpTBNcaEICdI11DhVmZ6Qxb-_GcuP8VbemKreHWNEu5id0ZviHPk7PAtLHqdBGr/pub?gid=1682849610&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01JPM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Australia
# https://fred.stlouisfed.org/series/IR3TIB01AUM156N
'rate_aus_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vS73ca8pMDndu3lH5SjmrIJS-HwWfDdqS2mh1YQkQwhGj3UtIauP12xjhmLusXag9ibZJE3YZsWLERT/pub?gid=1639615970&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01AUM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Canada
# https://fred.stlouisfed.org/series/IR3TIB01CAM156N
'rate_cnd_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vQ0Odb11l33qCVwPS6G2lxrUpfQ5DWXnGw8HFu6uUV_OUx7b-yBIQItN12TwLRTq3Bx3-fBe-pU86ve/pub?gid=483123093&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01CAM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for United Kingdom
# https://fred.stlouisfed.org/series/IR3TIB01GBM156N
'rate_gbp_3m_bank': {
'source': 'gsheet',
'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTtQSIEiSK1swVM_oElodv6YsjzojdxwfXZm-hDx68DQD6V3HjtuOYpHb4KUQC5uWDFoe9t09-Mibex/pub?gid=1120780951&single=true&output=csv",
'col_date': 'date',
'col_val': 'IR3TIB01GBM156N'.lower(),
'lag_days': 10,
},
},
"features": {
# "D1": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
# "W1": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
"D1": [],
"W1": [],
},
"features_val_map": {
"SMA10": ["val"],
"MACD": ["macd", "macdsignal", "macdhist"],
"ROC2": ["val"],
"MOM4": ["val"],
"RSI10": ["val"],
"BB20": ["upper","lower","mid"],
"CCI20": ["val"],
},
}
## Consolidator Settings
consolidator_settings = {
"D1": {
"timeframe_minutes": 24 * 60,
"consolidation_type": "quote",
# "indicators": [
# "SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
# ],
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
"W1": {
"timeframe_minutes": 7 * 24 * 60,
"consolidation_type": "quote",
# "indicators": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
}
## Indicators Settings
indicator_settings = {
"SMA10": {
"type": "SMA",
"lookback": 10,
"field": "Close",
"window": 3,
},
"MACD": {
"type": "MACD",
"window": 3,
},
"ROC2": {
"type": "ROC",
"lookback": 2,
"field": "Close",
"window": 3,
},
"MOM4": {
"type": "MOM",
"lookback": 2,
"field": "Close",
"window": 3,
},
"RSI10": {
"type": "RSI",
"lookback": 10,
"ma_type": "Simple",
"field": "Close",
"window": 3,
},
"BB20": {
"type": "BOLL",
"lookback": 20,
"ma_type": "Simple",
"std": 2,
"field": "Close",
"window": 3,
},
"CCI20": {
"type": "CCI",
"lookback": 20,
"field": "Close",
"window": 3,
},
"ATR10": {
"type": "ATR",
"lookback": 10,
"field": "Close",
"window": 3,
},
"ATR14": {
"type": "ATR",
"lookback": 14,
"field": "Close",
"window": 3,
},
"ATR21": {
"type": "ATR",
"lookback": 21,
"field": "Close",
"window": 3,
},
}
signal_settings = {
"FxLstm": {
##-##
"valid_tickers": ["EURUSD","USDJPY","USDCHF","GBPUSD","AUDUSD","USDCAD"],
##-##
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
},
}#region imports
from AlgorithmImports import *
#endregion
## General Settings
general_setting = {
"tickers": {
##-##
"EURUSD": {"type": "forex"},
# "USDJPY": {"type": "forex"},
# "USDCHF": {"type": "forex"},
# "GBPUSD": {"type": "forex"},
# "AUDUSD": {"type": "forex"},
# "USDCAD": {"type": "forex"},
##-##
},
"model_name": "ForexLSTM_V1_0",
"consolidator_timeframes": ["D1", "W1"],
##-##
"lstm_tickers": ['EURUSD'],
# "lstm_tickers": ['USDJPY'],
# "lstm_tickers": ['USDCHF'],
# "lstm_tickers": ['GBPUSD'],
# "lstm_tickers": ['AUDUSD'],
# "lstm_tickers": ['USDCAD'],
##-##
"fx_lstm_signal_name": 'FxLstm',
"order_counter_diff": 3,
"external_data": {},
# "external_data": {
# # SP500
# 'spy': {
# 'source': 'equity',
# 'ticker': 'SPY',
# },
# # Global X DAX Germany ETF
# 'dax': {
# 'source': 'equity',
# 'ticker': 'DAX',
# },
# # US Treasury
# 'us_treasury': {
# 'source': 'USTreasuryYieldCurveRate',
# 'ref': 'USTYCR',
# 'col_date': 'time',
# 'col_val': 'onemonth',
# },
# # Consumer Price Index for Inflation Rate
# # https://data.nasdaq.com/data/RATEINF-inflation-rates
# 'cpi_usa': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_USA",
# },
# 'cpi_eur': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_EUR",
# },
# 'cpi_deu': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_DEU",
# },
# # Federal Funds Effective Rate (DFF)
# # https://fred.stlouisfed.org/series/DFF
# 'dff': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vT5lyey5dhfrZifoZFuDwlQDOz6oILyUyAHTLVe2eqiLv9jWkNeIFITIeKqwBOtS8oEUOoZ2zXX1De7/pub?gid=1400614786&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'dff',
# 'lag_days': 2,
# },
# # Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for the Euro Area (19 Countries)
# # https://fred.stlouisfed.org/series/IRLTLT01EZM156N
# 'rate_eur_lt_gov': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSl_hxRnfcXnFly0Gh1vyZYNTRW6VTv-FQDlXuNUR1090RIst2a01nyhGl7tPR4VIcrgFfGBc3OSD72/pub?gid=1026565438&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IRLTLT01EZM156N'.lower(),
# 'lag_days': 2,
# },
# # Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for the Euro Area (19 Countries)
# # https://fred.stlouisfed.org/series/IR3TIB01EZM156N
# 'rate_eur_3m_bank': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vSkVfnj8N9AIsVF5PJN0JzU9ahw71nK_sTwY2qLKtNNxs1JI0STexUPEW15dY9bDUN8Fwql7_WUiKhK/pub?gid=2059310805&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IR3TIB01EZM156N'.lower(),
# 'lag_days': 2,
# },
# # Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for Germany
# # https://fred.stlouisfed.org/series/IRLTLT01DEM156N
# 'rate_deu_lt_gov': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vToUOn242L-w9ZWUXz_fU59aUc6oN5tDJEG8fu207zO7jMyfy5y7VesxH0mzEKaqwuU7WGOq7_xxDSu/pub?gid=2099864712&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IRLTLT01DEM156N'.lower(),
# 'lag_days': 2,
# },
# # Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Germany
# # https://fred.stlouisfed.org/series/IR3TIB01DEM156N
# 'rate_deu_3m_bank': {
# 'source': 'gsheet',
# 'link': "https://docs.google.com/spreadsheets/d/e/2PACX-1vTswIuhg3-tLwgP6RWSSPRyyLDpvHNqdlSgSNk91_SkUjKAD9_lyvhI84MAHRHzYdrIho1Narccx_w1/pub?gid=1568788544&single=true&output=csv",
# 'col_date': 'date',
# 'col_val': 'IR3TIB01DEM156N'.lower(),
# 'lag_days': 2,
# },
# },
"features": {
"D1": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10","PSAR",
# "ULTOSC","CHOP","DX14","PHASE","CRSI",
],
# "W1": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
# "D1": [],
"W1": [],
},
"features_val_map": {
"SMA10": ["val"],
"MACD": ["macd", "macdsignal", "macdhist"],
"ROC2": ["val"],
"MOM4": ["val"],
"RSI10": ["val"],
"BB20": ["upper","lower","mid"],
"CCI20": ["val"],
"ULTOSC": ["val"],
"CHOP": ["val"],
"DX14": ["val"],
"PHASE": ["val"],
"CRSI": ["val"],
"PSAR": ["val"],
},
}
## Consolidator Settings
consolidator_settings = {
"D1": {
"timeframe_minutes": 24 * 60,
"consolidation_type": "quote",
"indicators": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10",
"PSAR",
# "ULTOSC","CHOP","DX14","PHASE","CRSI",
],
# "indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
"W1": {
"timeframe_minutes": 7 * 24 * 60,
"consolidation_type": "quote",
# "indicators": ["SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20"],
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
}
## Indicators Settings
indicator_settings = {
"SMA10": {
"type": "SMA",
"lookback": 10,
"field": "Close",
"window": 3,
},
"MACD": {
"type": "MACD",
"window": 3,
},
"ROC2": {
"type": "ROC",
"lookback": 2,
"field": "Close",
"window": 3,
},
"MOM4": {
"type": "MOM",
"lookback": 2,
"field": "Close",
"window": 3,
},
"RSI10": {
"type": "RSI",
"lookback": 10,
"ma_type": "Simple",
"field": "Close",
"window": 3,
},
"BB20": {
"type": "BOLL",
"lookback": 20,
"ma_type": "Simple",
"std": 2,
"field": "Close",
"window": 3,
},
"CCI20": {
"type": "CCI",
"lookback": 20,
"field": "Close",
"window": 3,
},
"ATR10": {
"type": "ATR",
"lookback": 10,
"field": "Close",
"window": 3,
},
"ULTOSC": {
"type": "ULTOSC",
"window": 3,
},
"CHOP": {
"type": "CHOP",
"lookback": 52,
"window": 3,
},
"DX14": {
"type": "DX",
"lookback": 14,
"window": 3,
},
"PHASE": {
"type": "PHASE",
"lookback": 15,
"window": 3,
},
"CRSI": {
"type": "CRSI",
"rsi_len": 15,
"rsi_field": "Close",
"rsi_window": 21,
"window": 3,
},
"PSAR": {
"type": "PSAR",
"window": 3,
},
}
signal_settings = {
"FxLstm": {
##-##
"valid_tickers": ["EURUSD","USDJPY","USDCHF","GBPUSD","AUDUSD","USDCAD"],
##-##
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
},
}from AlgorithmImports import *
import numpy as np
from datetime import datetime, timedelta
from collections import deque
import talib as ta
class CustomSimpleMovingAverage:
def __init__(self, name, period):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.queue = deque(maxlen=period)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
# Update method is mandatory
def Update(self, EndTime, Val):
self.queue.appendleft(Val)
count = len(self.queue)
self.Time = EndTime
self.Value = sum(self.queue) / count
self.IsReady = count == self.queue.maxlen
def Undo(self):
del self.queue[0]
class CustomPhase:
def __init__(self, name, period):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.period = period
self.close = deque(maxlen=period)
self.high = deque(maxlen=period)
self.low = deque(maxlen=period)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.Time = bar.EndTime
RealPart = 0.0
ImagPart = 0.0
_count = min([len(self.close), len(self.high), len(self.low)])
for J in range(_count):
Weight = (
self.close[J] + self.close[J] + self.high[J] + self.low[J]
) * 10000
if self.period != 0:
RealPart = RealPart + np.cos(90 * J / self.period) * Weight * 2
ImagPart = (
(ImagPart + np.sin(90 * J / self.period) * Weight)
+ (ImagPart + np.sin(180 * J / self.period) * Weight)
) / 2
Phase = ((np.arctan(ImagPart / RealPart)) - 0.685) * 100
self.Value = Phase
self.IsReady = _count == self.period
class CustomCRSI:
def __init__(self, name, rsi_len, rsi_field, rsi_window):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.rsi_len = rsi_len
self.rsi_field = rsi_field
self.rsi_window = rsi_window
self.RSI = RelativeStrengthIndex(
f"{name}-RSI", rsi_len, MovingAverageType.Exponential
)
self.RSIval = deque(maxlen=rsi_window)
self.CRSIval = deque(maxlen=3)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.Time = bar.EndTime
self.RSI.Update(bar.EndTime, self.get_update_val(bar))
self.RSIval.appendleft(self.RSI.Current.Value)
vibration = 10
torque = 0.618 / (vibration + 1)
phasingLag = (vibration - 1) / 0.618
if len(self.RSIval) > int(phasingLag):
if len(self.CRSIval) > 1:
crsi1 = self.CRSIval[1]
else:
crsi1 = 0
self.Value = (
torque * (2 * self.RSIval[0] - self.RSIval[int(phasingLag)])
+ (1 - torque) * crsi1
)
else:
self.Value = 0
self.CRSIval.appendleft(self.Value)
self.IsReady = (
self.RSI.IsReady
and (len(self.RSIval) == self.rsi_window)
and (len(self.CRSIval) == 3)
)
def get_update_val(self, bar):
if self.rsi_field == "Close":
val = bar.Close
elif self.rsi_field == "High":
val = bar.High
elif self.rsi_field == "Low":
val = bar.Low
elif self.rsi_field == "Open":
val = bar.Open
return val
class CustomPivot:
def __init__(self, name, period):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.period = period
self.close = deque(maxlen=period)
self.high = deque(maxlen=period)
self.low = deque(maxlen=period)
self.arr_time = deque(maxlen=period)
self.p = deque(maxlen=period)
self.r1 = deque(maxlen=period)
self.s1 = deque(maxlen=period)
# self.r2 = deque(maxlen=period)
# self.s2 = deque(maxlen=period)
# self.r3 = deque(maxlen=period)
# self.s3 = deque(maxlen=period)
# self.r4 = deque(maxlen=period)
# self.s4 = deque(maxlen=period)
# self.r5 = deque(maxlen=period)
# self.s5 = deque(maxlen=period)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.Time = bar.EndTime
self.arr_time.appendleft(bar.EndTime)
pivotX_Median = (self.high[0] + self.low[0] + self.close[0]) / 3
self.Value = pivotX_Median
self.p.appendleft(pivotX_Median)
self.r1.appendleft(pivotX_Median * 2 - self.low[0])
self.s1.appendleft(pivotX_Median * 2 - self.high[0])
# self.r2.appendleft(pivotX_Median + 1 * (self.high[0] - self.low[0]))
# self.s2.appendleft(pivotX_Median - 1 * (self.high[0] - self.low[0]))
# self.r3.appendleft(pivotX_Median * 2 + (self.high[0] - 2 * self.low[0]))
# self.s3.appendleft(pivotX_Median * 2 - (2 * self.high[0] - self.low[0]))
# self.r4.appendleft(pivotX_Median * 3 + (self.high[0] - 3 * self.low[0]))
# self.s4.appendleft(pivotX_Median * 3 - (3 * self.high[0] - self.low[0]))
# self.r5.appendleft(pivotX_Median * 4 + (self.high[0] - 4 * self.low[0]))
# self.s5.appendleft(pivotX_Median * 4 - (4 * self.high[0] - self.low[0]))
self.IsReady = len(self.p) == self.period
class CustomChoppinessIndex:
def __init__(self, name, period):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.period = period
self.high = deque(maxlen=period)
self.low = deque(maxlen=period)
self.ATR = AverageTrueRange(f"{name}-RSI", 1)
self.ATRval = deque(maxlen=period)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.Time = bar.EndTime
self.ATR.Update(bar)
self.ATRval.appendleft(self.ATR.Current.Value)
if (max(self.high) - min(self.low)) == 0:
self.Value = 0
else:
self.Value = (
100
* np.log10(sum(self.ATRval) / (max(self.high) - min(self.low)))
/ np.log10(self.period)
)
self.IsReady = self.ATR.IsReady and (len(self.ATRval) == self.period)
class CustomDX:
def __init__(self, name, period):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.period = period
self.window_len = period * 2 + 25
self.high = deque(maxlen=self.window_len)
self.low = deque(maxlen=self.window_len)
self.close = deque(maxlen=self.window_len)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
ta_out = ta.DX(
np.array(self.high),
np.array(self.low),
np.array(self.close),
timeperiod=self.period,
)
self.Value = ta_out[-1]
else:
self.Value = 0
class CustomMACD:
def __init__(self, name):
self.Name = name
self.Time = datetime.min
self.IsReady = False
self.window_len = 100
self.close = deque(maxlen=self.window_len)
self.macd = 0
self.macdsignal = 0
self.macdhist = 0
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
macd, macdsignal, macdhist = ta.MACD(
np.array(self.close), fastperiod=12, slowperiod=26, signalperiod=9
)
self.macd = macd[-1]
self.macdsignal = macdsignal[-1]
self.macdhist = macdhist[-1]
else:
self.macd = 0
self.macdsignal = 0
self.macdhist = 0
class CustomULTOSC:
def __init__(self, name):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.window_len = 100
self.high = deque(maxlen=self.window_len)
self.low = deque(maxlen=self.window_len)
self.close = deque(maxlen=self.window_len)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
ta_out = ta.ULTOSC(
np.array(self.high),
np.array(self.low),
np.array(self.close),
timeperiod1=7,
timeperiod2=14,
timeperiod3=28,
)
self.Value = ta_out[-1]
else:
self.Value = 0
## Version
# Forex LSTM V 1.0
from AlgorithmImports import *
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from collections import deque
import pickle
from config import (
general_setting,
consolidator_settings,
indicator_settings,
signal_settings,
)
from data_classes import (
SymbolData,
MarketHours,
)
from signal_classes import (
FxLstmSignal,
)
signal_mapping = {
"FxLstm_Both_EURUSD": FxLstmSignal,
"FxLstm_Both_EURUSD_Trail": FxLstmSignal,
"FxLstm_Hybrid_EURUSD": FxLstmSignal,
"FxLstm_Hybrid_EURUSD_Trail": FxLstmSignal,
"FxLstm_Both_USDJPY": FxLstmSignal,
"FxLstm_Both_USDJPY_Trail": FxLstmSignal,
"FxLstm_Hybrid_GBPUSD": FxLstmSignal,
"FxLstm_Hybrid_GBPUSD_Trail": FxLstmSignal,
"FxLstm_Both_AUDUSD": FxLstmSignal,
"FxLstm_Both_AUDUSD_Trail": FxLstmSignal,
"FxLstm_Hybrid_AUDUSD": FxLstmSignal,
"FxLstm_Hybrid_AUDUSD_Trail": FxLstmSignal,
}
from QuantConnect.DataSource import *
class FxLstmAlgo(QCAlgorithm):
def Initialize(self):
## For Optimization
# _signal = ["FxLstm_Both_Long","FxLstm_Both_Short","FxLstm_Hybrid_Long","FxLstm_Hybrid_Short"][3]
# _signal = ["FxLstm_Both","FxLstm_Hybrid"][0]
# general_setting["signals"] = [_signal]
# StopMultiplier = [0.1, 0.2, 0.25, 0.5, 0.75, 1.0][int(self.GetParameter('multiplier'))]
# signal_settings[_signal]["longStopMultiplier"] = StopMultiplier
# signal_settings[_signal]["shortStopMultiplier"] = StopMultiplier
# sma_filter_lookback = [10, 20, 50, 100, 200][int(self.GetParameter('multiplier'))]
# for _signal in ["FxLstm_Both_Long","FxLstm_Both_Short","FxLstm_Hybrid_Long","FxLstm_Hybrid_Short"]:
# signal_settings[_signal]["sma_filter_lookback"] = sma_filter_lookback
# _signal = [
# "FxLstm_Both_EURUSD",
# "FxLstm_Hybrid_EURUSD",
# "FxLstm_Both_USDJPY",
# "FxLstm_Hybrid_GBPUSD",
# "FxLstm_Both_AUDUSD",
# "FxLstm_Hybrid_AUDUSD"
# ][5]
# # if int(self.GetParameter('multiplier')) == 0:
# # signal_settings[_signal]["atrLength"] = 10
# # if int(self.GetParameter('multiplier')) == 1:
# # signal_settings[_signal]["atrLength"] = 14
# # if int(self.GetParameter('multiplier')) == 2:
# # signal_settings[_signal]["atrLength"] = 21
# # StopMultiplier = [0.1, 0.2, 0.25, 0.5, 0.75, 1.0][int(self.GetParameter('multiplier'))]
# # signal_settings[_signal]["longStopMultiplier"] = StopMultiplier
# # signal_settings[_signal]["shortStopMultiplier"] = StopMultiplier
# # RiskRewardMultiplier = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0][int(self.GetParameter('multiplier'))]
# # signal_settings[_signal]["longRiskRewardMultiplier"] = RiskRewardMultiplier
# # signal_settings[_signal]["shortRiskRewardMultiplier"] = RiskRewardMultiplier
# trailStopSize = [0.1, 0.2, 0.5, 0.75, 1.0, 1.25, 1.5, 2.0, 2.5, 3.0, 4.0, 5.0][int(self.GetParameter('multiplier'))]
# signal_settings[_signal]["useTralingStop"] = True
# signal_settings[_signal]["trailStopSize"] = trailStopSize
## Full Baseline
# SR: 1.122, PSR: 58.145%
# SR: 1.172, PSR: 63.188%
# SR: 1.205, PSR: 66.028%
# SR: 1.215, PSR: 64.886%
# SR: 1.270, PSR: 69.364%
# SR: 1.320, PSR: 73.606%
# SR: 1.305, PSR: 71.873%
# SR: 1.317, PSR: 72.794%
# SR: 1.343, PSR: 73.131%
##
self.SetTimeZone(TimeZones.Johannesburg)
##-##
self.SetStartDate(2018, 1, 1)
##-##
self.SetCash(10000)
self.SetWarmUp(int(12 * 20 * 24 * 60), Resolution.Minute)
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
self.to_plot = False
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.signal_settings = signal_settings
self.ref_ticker = "EURUSD"
##-##
self.prediction_dict_full_TEMP = {}
for lstm_ticker in self.general_setting["lstm_tickers"]:
self.prediction_dict_full_TEMP[lstm_ticker] = pickle.loads(bytes(self.ObjectStore.ReadBytes(f"PREDICTIONS_{lstm_ticker}")))
##-##
self.prediction_dict = {}
for _signal in self.general_setting["signals"]:
self.prediction_dict[_signal] = {}
for ticker in self.general_setting["lstm_tickers"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
self.prediction_dict[_signal][ticker] = 1
# Data Initialization
self.Data = {}
self.Signal = {}
self.Counter = {}
self.SymbolMarketHours = {}
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.data_list = []
self.output_data_dict = {}
for ticker in self.general_setting["tickers"]:
if general_setting["tickers"][ticker]["type"] == "equity":
symbol = self.AddEquity(
ticker,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.Raw,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "forex":
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "cfd":
symbol = self.AddCfd(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.Data[symbol] = SymbolData(
self,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
)
self.Counter[symbol] = {}
self.Counter[symbol]["counter"] = 0
self.Counter[symbol]["last_order_counter"] = 0
self.SymbolMarketHours[symbol] = MarketHours(self, symbol)
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
self.Signal[symbol] = {}
for _signal in self.general_setting["signals"]:
self.Signal[symbol][_signal] = signal_mapping[_signal](
self, symbol, ticker, self.general_setting, self.signal_settings[_signal]
)
self.Schedule.On(
self.DateRules.MonthEnd(self.ref_ticker),
self.TimeRules.BeforeMarketClose(self.ref_ticker, 0),
self.SaveData,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker),
self.SOD,
)
##-##
# self.external_data = {}
# for _dn in self.general_setting["external_data"]:
# self.external_data[_dn] = {}
# self.external_data[_dn]['time'] = None
# self.external_data[_dn]['value'] = None
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'gsheet':
# link = self.general_setting["external_data"][_dn]['link']
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# data = self.Download(link)
# rows = []
# for row in data.split('\n'):
# rows.append(row.replace('\r','').lower().split(','))
# data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
# data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
# data_df[col_val] = data_df[col_val].astype(float)
# self.external_data[_dn]['data'] = data_df.copy()
# if source == 'NasdaqDataLink':
# ref = self.general_setting["external_data"][_dn]['ref']
# self.external_data[_dn]['symbol'] = self.AddData(NasdaqDataLink, ref, Resolution.Daily).Symbol
# if source == 'equity':
# ticker = self.general_setting["external_data"][_dn]['ticker']
# self.external_data[_dn]['symbol'] = self.AddEquity(ticker, Resolution.Daily).Symbol
# if source == 'USTreasuryYieldCurveRate':
# ref = self.general_setting["external_data"][_dn]['ref']
# self.external_data[_dn]['symbol'] = self.AddData(USTreasuryYieldCurveRate, ref).Symbol
##-##
def SOD(self):
##-##
pass
# for _dn in self.general_setting["external_data"]:
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'gsheet':
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# lag_days = self.general_setting["external_data"][_dn]['lag_days']
# data = self.external_data[_dn]['data'][self.external_data[_dn]['data'][col_date] < (self.Time - timedelta(days=lag_days))]
# if len(data) > 0:
# self.external_data[_dn]['time'] = data[col_date].values[-1]
# self.external_data[_dn]['value'] = data[col_val].values[-1]
# if source == 'USTreasuryYieldCurveRate':
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# symbol = self.external_data[_dn]['symbol']
# history = self.History(USTreasuryYieldCurveRate, symbol, 1, Resolution.Daily)
# history = history.reset_index()
# if len(history) > 0:
# if col_val in history.columns:
# self.external_data[_dn]['time'] = pd.to_datetime(history[col_date], utc=True).iloc[0].replace(tzinfo=None)
# self.external_data[_dn]['value'] = history[col_val].values[0]
##-##
def SaveData(self):
##-##
pass
# self.output_data_dict["data"] = self.data_list
# self.output_data_dict["version"] = "2"
# # self.ObjectStore.SaveBytes("FUNDAMENTAL_DATA", pickle.dumps(self.output_data_dict))
# self.ObjectStore.SaveBytes("TECHNICAL_DATA", pickle.dumps(self.output_data_dict))
##-##
def OnData(self, data):
##-##
# for _dn in self.general_setting["external_data"]:
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'NasdaqDataLink':
# symbol = self.external_data[_dn]['symbol']
# if data.ContainsKey(symbol):
# self.external_data[_dn]['value'] = data[symbol].Value
# if source == 'equity':
# symbol = self.external_data[_dn]['symbol']
# if data.ContainsKey(symbol):
# if data[symbol] is not None:
# self.external_data[_dn]['time'] = data[symbol].Time
# self.external_data[_dn]['value'] = data[symbol].Price
##-##
FxLstm_SymbolQuantity = {}
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
ticker = self.symbol_ticker_map[symbol]
is_valid_time = self.Time.minute == 0
is_valid_time = is_valid_time and (self.Time.hour in [self.general_setting['FxLstm_prediction_hour']])
if is_valid_time:
if ticker in self.general_setting["lstm_tickers"]:
##-##
# symbol = self.ticker_symbol_map[ticker]
# symbolData = self.Data[symbol]
# if not (
# data.ContainsKey(symbol)
# and data[symbol] is not None
# and symbolData.IsReady
# ):
# continue
# data_dict = {}
# data_dict["datetime"] = self.Time
# data_dict["ticker"] = ticker
# data_dict["symbol"] = symbol.Value
# data_dict["isWarmingUp"] = self.IsWarmingUp
# data_dict["symbolTime"] = str(data[symbol].Time)
# data_dict["price"] = np.round(data[symbol].Price, 6)
# # Daily Data
# _consolidator = symbolData.consolidators["D1"]
# data_dict["close_D1"] = _consolidator.close[0]
# # External Data
# for _dn in self.general_setting["external_data"]:
# data_dict[f"{_dn}_time"] = self.external_data[_dn]['time']
# data_dict[_dn] = self.external_data[_dn]['value']
# # Technical Features
# for _tf in self.general_setting["features"]:
# _consolidator = symbolData.consolidators[_tf]
# for _in in self.general_setting["features"][_tf]:
# _indicator = _consolidator.indicators[_in]
# if _in in self.general_setting["features_val_map"]:
# for _v in self.general_setting["features_val_map"][
# _in
# ]:
# data_dict[f"{_tf}-{_in}-{_v}"] = np.round(
# _indicator[_v][0], 5
# )
# # if not self.IsWarmingUp:
# # self.data_list += [data_dict]
# self.data_list += [data_dict]
##-##
for _signal in self.general_setting["signals"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
pred_type = self.signal_settings[_signal]['pred_type']
lstm_ticker = self.signal_settings[_signal]['lstm_ticker']
if self.Time in self.prediction_dict_full_TEMP[lstm_ticker][pred_type]:
self.prediction_dict[_signal][ticker] = self.prediction_dict_full_TEMP[lstm_ticker][pred_type][self.Time]
self.Signal[symbol][_signal].update_prediction_direction(self.prediction_dict[_signal][ticker])
symbolQuantity = 0
for _signal in self.general_setting["signals"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
to_exit = self.Signal[symbol][_signal].check_exit(symbolData, data[symbol].Price, data[symbol].Time)
if to_exit:
self.Signal[symbol][_signal].update_exit()
has_enter = self.Signal[symbol][_signal].enter(symbolData, data[symbol].Price, data[symbol].Time)
quantity = self.Signal[symbol][_signal].quantity * self.Signal[symbol][_signal].allocation_multiplier
quantity = int(np.ceil(quantity))
symbolQuantity += quantity
FxLstm_SymbolQuantity[symbol] = symbolQuantity
## Aggregate symbol quantities across strategies
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
self.Counter[symbol]["counter"] += 1
symbolQuantity = 0
if symbol in FxLstm_SymbolQuantity:
symbolQuantity += FxLstm_SymbolQuantity[symbol]
if not self.IsWarmingUp:
# In case orders takes longer than 1 bar to be filled. Only send market orders every 3 minutes
if (self.Counter[symbol]["counter"] - self.Counter[symbol]["last_order_counter"]) >= self.general_setting["order_counter_diff"]:
if (symbolQuantity - self.Portfolio[symbol].Quantity) != 0:
self.MarketOrder(symbol, symbolQuantity - self.Portfolio[symbol].Quantity)
self.Counter[symbol]["last_order_counter"] = self.Counter[symbol]["counter"]
## Version
# Forex LSTM V 1.0
from AlgorithmImports import *
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from collections import deque
import pickle
from config import (
general_setting,
consolidator_settings,
indicator_settings,
signal_settings,
)
from data_classes import (
SymbolData,
MarketHours,
)
from signal_classes import (
FxLstmSignal,
)
signal_mapping = {
"FxLstm_Both": FxLstmSignal,
"FxLstm_Hybrid": FxLstmSignal,
"FxLstm_Both_Long": FxLstmSignal,
"FxLstm_Both_Short": FxLstmSignal,
"FxLstm_Hybrid_Long": FxLstmSignal,
"FxLstm_Hybrid_Short": FxLstmSignal,
}
from QuantConnect.DataSource import *
class FxLstmAlgo(QCAlgorithm):
def Initialize(self):
## For Optimization
# _signal = ["FxLstm_Both_Long","FxLstm_Both_Short","FxLstm_Hybrid_Long","FxLstm_Hybrid_Short"][3]
# _signal = ["FxLstm_Both","FxLstm_Hybrid"][0]
# general_setting["signals"] = [_signal]
# StopMultiplier = [0.1, 0.2, 0.25, 0.5, 0.75, 1.0][int(self.GetParameter('multiplier'))]
# signal_settings[_signal]["longStopMultiplier"] = StopMultiplier
# signal_settings[_signal]["shortStopMultiplier"] = StopMultiplier
# sma_filter_lookback = [10, 20, 50, 100, 200][int(self.GetParameter('multiplier'))]
# for _signal in ["FxLstm_Both_Long","FxLstm_Both_Short","FxLstm_Hybrid_Long","FxLstm_Hybrid_Short"]:
# signal_settings[_signal]["sma_filter_lookback"] = sma_filter_lookback
# _signal = ["FxLstm_Both","FxLstm_Hybrid"][1]
# if int(self.GetParameter('multiplier')) == 0:
# signal_settings[f"{_signal}"]["longRiskRewardMultiplier"] = 2
# signal_settings[f"{_signal}"]["shortRiskRewardMultiplier"] = 3
# if int(self.GetParameter('multiplier')) == 1:
# signal_settings[f"{_signal}"]["longRiskRewardMultiplier"] = 3
# signal_settings[f"{_signal}"]["shortRiskRewardMultiplier"] = 2
# if int(self.GetParameter('multiplier')) == 2:
# signal_settings[f"{_signal}"]["longRiskRewardMultiplier"] = 2
# signal_settings[f"{_signal}"]["shortRiskRewardMultiplier"] = 2
# if int(self.GetParameter('multiplier')) == 3:
# signal_settings[f"{_signal}"]["longRiskRewardMultiplier"] = 3
# signal_settings[f"{_signal}"]["shortRiskRewardMultiplier"] = 4
# if int(self.GetParameter('multiplier')) == 4:
# signal_settings[f"{_signal}"]["longRiskRewardMultiplier"] = 4
# signal_settings[f"{_signal}"]["shortRiskRewardMultiplier"] = 3
# if int(self.GetParameter('multiplier')) == 5:
# signal_settings[f"{_signal}"]["longRiskRewardMultiplier"] = 4
# signal_settings[f"{_signal}"]["shortRiskRewardMultiplier"] = 4
## 1.053
# FxLstm_Both_Long
# 0.5
# FxLstm_Both_Short
# 0.2
# FxLstm_Hybrid_Long
# 1.0
# FxLstm_Hybrid_Short
# 0.5
## Baseline: 0.542
## SMA 10 Long: 0.537
## SMA 20 Long: 0.537
## SMA 50 Long: 0.542
## SMA 10 / 50: 0.742
## SMA 10 / 100: 0.742
## SMA 10 / 200: 0.742
## SMA 20 / 50:
# FxLstm_Both: SR: 0.512
# FxLstm_Hybrid: SR: 0.22
## Both: SR: 0.421
# SMA5: 0.457
# SMA20: 0.561
# SMA50: 0.572
# SMA100: 0.952
# SMA200: 0.852
# SMA 5/10: 0.421
# SMA 5/100: 0.474
# SMA 5/200: 0.76
# SMA 10/20: 0.311
# SMA 10/50: 0.519
# SMA 10/100: 0.431
# SMA 10/200: 0.532
# SMA 20/50: 0.646
# SMA 20/100: 0.459
# RiskRewardMultiplier = [0.5, 0.75, 1.0, 1.2, 1.25, 1.5, 1.75, 2.0, 2.5, 3.0, 4.0, 5.0, 6.0, 7.0, 10.0][int(self.GetParameter('multiplier'))]
# signal_settings["FxLstm"]["longRiskRewardMultiplier"] = RiskRewardMultiplier
# signal_settings["FxLstm"]["shortRiskRewardMultiplier"] = RiskRewardMultiplier
# trailStopSize = [0.1, 0.2, 0.25, 0.5, 0.75, 1.0, 1.2, 1.25, 1.5, 1.75, 2.0, 2.5, 3.0, 4.0, 5.0][int(self.GetParameter('multiplier'))]
# signal_settings["FxLstm"]["trailStopSize"] = trailStopSize
##
self.SetTimeZone(TimeZones.Johannesburg)
##-##
self.SetStartDate(2018, 1, 1)
##-##
self.SetCash(10000)
##-##
self.SetWarmUp(int(12 * 20 * 24 * 60), Resolution.Minute)
# self.SetWarmUp(int(1 * 20 * 24 * 60), Resolution.Minute)
##-##
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
self.to_plot = False
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.signal_settings = signal_settings
self.ref_ticker = "EURUSD"
##-##
self.prediction_dict_full_TEMP = pickle.loads(bytes(self.ObjectStore.ReadBytes(f"PREDICTIONS")))
##-##
self.prediction_dict = {}
for _signal in self.general_setting["signals"]:
self.prediction_dict[_signal] = {}
for ticker in self.general_setting["lstm_tickers"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
self.prediction_dict[_signal][ticker] = 1
self.threshold_dict = {}
for ticker in self.general_setting["lstm_tickers"]:
self.threshold_dict[ticker] = 0
# Data Initialization
self.Data = {}
self.Signal = {}
self.Counter = {}
self.SymbolMarketHours = {}
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.data_list = []
self.output_data_dict = {}
for ticker in self.general_setting["tickers"]:
if general_setting["tickers"][ticker]["type"] == "equity":
symbol = self.AddEquity(
ticker,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.Raw,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "forex":
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "cfd":
symbol = self.AddCfd(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.Data[symbol] = SymbolData(
self,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
)
self.Counter[symbol] = {}
self.Counter[symbol]["counter"] = 0
self.Counter[symbol]["last_order_counter"] = 0
self.SymbolMarketHours[symbol] = MarketHours(self, symbol)
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
self.Signal[symbol] = {}
for _signal in self.general_setting["signals"]:
self.Signal[symbol][_signal] = signal_mapping[_signal](
self, symbol, ticker, self.general_setting, self.signal_settings[_signal]
)
self.Schedule.On(
self.DateRules.MonthEnd(self.ref_ticker),
self.TimeRules.BeforeMarketClose(self.ref_ticker, 0),
self.SaveData,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker),
self.SOD,
)
##-##
# self.external_data = {}
# for _dn in self.general_setting["external_data"]:
# self.external_data[_dn] = {}
# self.external_data[_dn]['time'] = None
# self.external_data[_dn]['value'] = None
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'gsheet':
# link = self.general_setting["external_data"][_dn]['link']
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# data = self.Download(link)
# rows = []
# for row in data.split('\n'):
# rows.append(row.replace('\r','').lower().split(','))
# data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
# data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
# data_df[col_val] = data_df[col_val].astype(float)
# self.external_data[_dn]['data'] = data_df.copy()
# if source == 'NasdaqDataLink':
# ref = self.general_setting["external_data"][_dn]['ref']
# self.external_data[_dn]['symbol'] = self.AddData(NasdaqDataLink, ref, Resolution.Daily).Symbol
# if source == 'equity':
# ticker = self.general_setting["external_data"][_dn]['ticker']
# self.external_data[_dn]['symbol'] = self.AddEquity(ticker, Resolution.Daily).Symbol
# if source == 'USTreasuryYieldCurveRate':
# ref = self.general_setting["external_data"][_dn]['ref']
# self.external_data[_dn]['symbol'] = self.AddData(USTreasuryYieldCurveRate, ref).Symbol
##-##
def SOD(self):
##-##
pass
# for _dn in self.general_setting["external_data"]:
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'gsheet':
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# lag_days = self.general_setting["external_data"][_dn]['lag_days']
# data = self.external_data[_dn]['data'][self.external_data[_dn]['data'][col_date] < (self.Time - timedelta(days=lag_days))]
# if len(data) > 0:
# self.external_data[_dn]['time'] = data[col_date].values[-1]
# self.external_data[_dn]['value'] = data[col_val].values[-1]
# if source == 'USTreasuryYieldCurveRate':
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# symbol = self.external_data[_dn]['symbol']
# history = self.History(USTreasuryYieldCurveRate, symbol, 1, Resolution.Daily)
# history = history.reset_index()
# if len(history) > 0:
# if col_val in history.columns:
# self.external_data[_dn]['time'] = pd.to_datetime(history[col_date], utc=True).iloc[0].replace(tzinfo=None)
# self.external_data[_dn]['value'] = history[col_val].values[0]
##-##
def SaveData(self):
##-##
pass
# self.output_data_dict["data"] = self.data_list
# self.output_data_dict["version"] = "2"
# # self.ObjectStore.SaveBytes("FUNDAMENTAL_DATA", pickle.dumps(self.output_data_dict))
# self.ObjectStore.SaveBytes("TECHNICAL_DATA", pickle.dumps(self.output_data_dict))
##-##
def OnData(self, data):
##-##
# for _dn in self.general_setting["external_data"]:
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'NasdaqDataLink':
# symbol = self.external_data[_dn]['symbol']
# if data.ContainsKey(symbol):
# self.external_data[_dn]['value'] = data[symbol].Value
# if source == 'equity':
# symbol = self.external_data[_dn]['symbol']
# if data.ContainsKey(symbol):
# if data[symbol] is not None:
# self.external_data[_dn]['time'] = data[symbol].Time
# self.external_data[_dn]['value'] = data[symbol].Price
##-##
FxLstm_SymbolQuantity = {}
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
ticker = self.symbol_ticker_map[symbol]
is_valid_time = self.Time.minute == 0
is_valid_time = is_valid_time and (self.Time.hour in [self.general_setting['FxLstm_prediction_hour']])
if is_valid_time:
if ticker in self.general_setting["lstm_tickers"]:
##-##
# symbol = self.ticker_symbol_map[ticker]
# symbolData = self.Data[symbol]
# if not (
# data.ContainsKey(symbol)
# and data[symbol] is not None
# and symbolData.IsReady
# ):
# continue
# data_dict = {}
# data_dict["datetime"] = self.Time
# data_dict["ticker"] = ticker
# data_dict["symbol"] = symbol.Value
# data_dict["isWarmingUp"] = self.IsWarmingUp
# data_dict["symbolTime"] = str(data[symbol].Time)
# data_dict["price"] = np.round(data[symbol].Price, 6)
# # Daily Data
# _consolidator = symbolData.consolidators["D1"]
# data_dict["close_D1"] = _consolidator.close[0]
# # External Data
# for _dn in self.general_setting["external_data"]:
# data_dict[f"{_dn}_time"] = self.external_data[_dn]['time']
# data_dict[_dn] = self.external_data[_dn]['value']
# # Technical Features
# for _tf in self.general_setting["features"]:
# _consolidator = symbolData.consolidators[_tf]
# for _in in self.general_setting["features"][_tf]:
# _indicator = _consolidator.indicators[_in]
# if _in in self.general_setting["features_val_map"]:
# for _v in self.general_setting["features_val_map"][
# _in
# ]:
# data_dict[f"{_tf}-{_in}-{_v}"] = np.round(
# _indicator[_v][0], 5
# )
# # if not self.IsWarmingUp:
# # self.data_list += [data_dict]
# self.data_list += [data_dict]
##-##
if self.Time in self.prediction_dict_full_TEMP['threshold']:
self.threshold_dict[ticker] = self.prediction_dict_full_TEMP['threshold'][self.Time]
for _signal in self.general_setting["signals"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
pred_type = self.signal_settings[_signal]['pred_type']
if self.Time in self.prediction_dict_full_TEMP[pred_type]:
self.prediction_dict[_signal][ticker] = self.prediction_dict_full_TEMP[pred_type][self.Time]
self.Signal[symbol][_signal].update_prediction_direction(self.prediction_dict[_signal][ticker])
self.Signal[symbol][_signal].update_prediction_threshold(self.threshold_dict[ticker])
symbolQuantity = 0
for _signal in self.general_setting["signals"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
to_exit = self.Signal[symbol][_signal].check_exit(symbolData, data[symbol].Price, data[symbol].Time)
if to_exit:
self.Signal[symbol][_signal].update_exit()
has_enter = self.Signal[symbol][_signal].enter(symbolData, data[symbol].Price, data[symbol].Time)
quantity = self.Signal[symbol][_signal].quantity * self.Signal[symbol][_signal].allocation_multiplier
quantity = int(np.ceil(quantity))
symbolQuantity += quantity
FxLstm_SymbolQuantity[symbol] = symbolQuantity
## Aggregate symbol quantities across strategies
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
self.Counter[symbol]["counter"] += 1
symbolQuantity = 0
if symbol in FxLstm_SymbolQuantity:
symbolQuantity += FxLstm_SymbolQuantity[symbol]
if not self.IsWarmingUp:
# In case orders takes longer than 1 bar to be filled. Only send market orders every 3 minutes
if (self.Counter[symbol]["counter"] - self.Counter[symbol]["last_order_counter"]) >= self.general_setting["order_counter_diff"]:
if (symbolQuantity - self.Portfolio[symbol].Quantity) != 0:
self.MarketOrder(symbol, symbolQuantity - self.Portfolio[symbol].Quantity)
self.Counter[symbol]["last_order_counter"] = self.Counter[symbol]["counter"]
## Version
# Forex LSTM V 1.0
##-##
IS_LIVE = True
# TO_SAVE_DATA = False
if IS_LIVE:
TO_SAVE_DATA = False
else:
TO_SAVE_DATA = True
##-##
from AlgorithmImports import *
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from collections import deque
import pickle
import torch
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(f"device: {device}")
from config import (
general_setting,
consolidator_settings,
indicator_settings,
signal_settings,
model_settings,
)
from data_classes import (
SymbolData,
MarketHours,
)
from signal_classes import (
FxLstmSignal,
)
from model_functions import (
get_threshold,
set_seed,
)
from model_classes_both import get_torch_rnn_dataloaders as get_torch_rnn_dataloaders_both
from model_classes_both import get_rnn_model as get_rnn_model_both
from model_classes_both import get_predictions as get_predictions_both
from model_classes_hybrid import get_torch_rnn_dataloaders as get_torch_rnn_dataloaders_hybrid
from model_classes_hybrid import get_rnn_model as get_rnn_model_hybrid
from model_classes_hybrid import get_predictions as get_predictions_hybrid
from model_classes_hybrid import get_regression_pred_decision, get_prediction_hybrid_regression
from sklearn.metrics import mean_squared_error
signal_mapping = {
"FxLstm_Both_EURUSD": FxLstmSignal,
"FxLstm_Both_EURUSD_Trail": FxLstmSignal,
"FxLstm_Hybrid_EURUSD": FxLstmSignal,
"FxLstm_Hybrid_EURUSD_Trail": FxLstmSignal,
"FxLstm_Both_USDJPY": FxLstmSignal,
"FxLstm_Both_USDJPY_Trail": FxLstmSignal,
"FxLstm_Hybrid_GBPUSD": FxLstmSignal,
"FxLstm_Hybrid_GBPUSD_Trail": FxLstmSignal,
"FxLstm_Both_AUDUSD": FxLstmSignal,
"FxLstm_Both_AUDUSD_Trail": FxLstmSignal,
"FxLstm_Hybrid_AUDUSD": FxLstmSignal,
"FxLstm_Hybrid_AUDUSD_Trail": FxLstmSignal,
}
from QuantConnect.DataSource import *
class FxLstmAlgo(QCAlgorithm):
def Initialize(self):
self.SetTimeZone(TimeZones.Johannesburg)
if TO_SAVE_DATA:
self.SetStartDate(2013, 1, 1) ## SR: 0.574, 0.811, 0.714
# self.SetEndDate(2017, 12, 31)
# self.SetEndDate(2018, 12, 31)
# self.SetEndDate(2020, 12, 31)
# self.SetEndDate(2021, 12, 31)
else:
# self.SetStartDate(2018, 1, 1) ## SR: 0.799, 1.094, 1.16, 1.054, 1.078, 1.097, 1.087
# self.SetStartDate(2019, 1, 1) ## SR: 0.573, 0.908, 0.986, 0.997, 1.028, 1.076
# self.SetStartDate(2021, 1, 1) ## SR: 0.489, 1.108, 1.256, 1.333, 1.333, 1.374
self.SetStartDate(2022, 1, 1) ## SR: 0.783, 1.49, 1.729, 1.674, 1.674, 1.696, 1.708
self.SetCash(10000)
if TO_SAVE_DATA:
self.SetWarmUp(int(12 * 20 * 24 * 60), Resolution.Minute)
else:
self.SetWarmUp(int(6 * 20 * 24 * 60), Resolution.Minute)
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
self.to_plot = False
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.signal_settings = signal_settings
self.model_settings = model_settings
self.ref_ticker = "EURUSD"
self.model_name = general_setting["model_name"]
self.month_start_date = None
self.IS_LIVE = IS_LIVE
self.TO_SAVE_DATA = TO_SAVE_DATA
self.prediction_dict = {}
for _signal in self.general_setting["signals"]:
self.prediction_dict[_signal] = {}
for ticker in self.general_setting["lstm_tickers"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
self.prediction_dict[_signal][ticker] = 1
# Data Initialization
self.Data = {}
self.Signal = {}
self.Counter = {}
self.SymbolMarketHours = {}
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.output_data_dict = {}
for ticker in self.general_setting["tickers"]:
if general_setting["tickers"][ticker]["type"] == "equity":
symbol = self.AddEquity(
ticker,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.Raw,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "forex":
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "cfd":
symbol = self.AddCfd(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.Data[symbol] = SymbolData(
self,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
)
self.Counter[symbol] = {}
self.Counter[symbol]["counter"] = 0
self.Counter[symbol]["last_order_counter"] = 0
self.SymbolMarketHours[symbol] = MarketHours(self, symbol)
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
self.Signal[symbol] = {}
for _signal in self.general_setting["signals"]:
self.Signal[symbol][_signal] = signal_mapping[_signal](
self, symbol, ticker, self.general_setting, self.signal_settings[_signal]
)
# Model Initialization
self.Models = {}
self.Scalers = {}
self.ModelParams = {}
for lstm_ticker in self.general_setting["lstm_tickers"]:
self.Models[lstm_ticker] = {}
self.Scalers[lstm_ticker] = {}
self.ModelParams[lstm_ticker] = {}
for model_type in self.general_setting["model_types"]:
self.Models[lstm_ticker][model_type] = {}
self.Scalers[lstm_ticker][model_type] = {}
self.ModelParams[lstm_ticker][model_type] = {}
if model_type == 'both':
self.Models[lstm_ticker][model_type]['both'] = None
self.Scalers[lstm_ticker][model_type]['both'] = None
if model_type == 'hybrid':
self.Models[lstm_ticker][model_type]['fundamental'] = None
self.Models[lstm_ticker][model_type]['technical'] = None
self.Scalers[lstm_ticker][model_type]['fundamental'] = None
self.Scalers[lstm_ticker][model_type]['technical'] = None
# Model Data Initialization
self.data_list_tickers = {}
self.has_initialized_model_data = {}
if self.TO_SAVE_DATA:
self.ModelData = {}
else:
self.ModelData = pickle.loads(bytes(self.ObjectStore.ReadBytes(f"MODEL_DATA_{self.model_name}")))
for lstm_ticker in self.general_setting["lstm_tickers"]:
self.data_list_tickers[lstm_ticker] = []
self.has_initialized_model_data[lstm_ticker] = False
if lstm_ticker not in self.ModelData:
self.ModelData[lstm_ticker] = pd.DataFrame()
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker),
self.Start_Of_Day,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.At(self.general_setting['FxLstm_prediction_hour'], 5, 0),
self.Prepare_Model_Data,
)
self.Schedule.On(
self.DateRules.MonthStart(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker, 1),
self.Get_Month_Start_Date,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker, 2),
self.Train_Model_Both,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker, 3),
self.Train_Model_Hybrid,
)
self.external_data = {}
for _dn in self.general_setting["external_data"]:
self.external_data[_dn] = {}
self.external_data[_dn]['time'] = None
self.external_data[_dn]['value'] = None
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
self.Log(f"{str(self.Time)}: {_dn}: Loading Initial GSheet Data")
link = self.general_setting["external_data"][_dn]['link']
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
to_run = True
while to_run:
try:
data = self.Download(link)
rows = []
for row in data.split('\n'):
rows.append(row.replace('\r', '').lower().split(','))
data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
data_df[col_val] = data_df[col_val].astype(float)
self.external_data[_dn]['data'] = data_df.copy()
to_run = False
except:
pass
self.Log(f"{str(self.Time)}: {_dn}: Initial GSheet Data Loaded")
if source == 'NasdaqDataLink':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(NasdaqDataLink, ref, Resolution.Daily).Symbol
if source == 'equity':
ticker = self.general_setting["external_data"][_dn]['ticker']
self.external_data[_dn]['symbol'] = self.AddEquity(ticker, Resolution.Daily).Symbol
if source == 'USTreasuryYieldCurveRate':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(USTreasuryYieldCurveRate, ref).Symbol
def Start_Of_Day(self):
if self.IS_LIVE and (not self.IsWarmingUp):
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
self.Log(f"{str(self.Time)}: {_dn}: Loading GSheet Data")
link = self.general_setting["external_data"][_dn]['link']
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
to_run = True
while to_run:
try:
data = self.Download(link)
rows = []
for row in data.split('\n'):
rows.append(row.replace('\r', '').lower().split(','))
data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
data_df[col_val] = data_df[col_val].astype(float)
self.external_data[_dn]['data'] = data_df.copy()
to_run = False
except:
pass
self.Log(f"{str(self.Time)}: {_dn}: GSheet Data Loaded")
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
lag_days = self.general_setting["external_data"][_dn]['lag_days']
data = self.external_data[_dn]['data'][self.external_data[_dn]['data'][col_date] < (self.Time - timedelta(days=lag_days))]
if len(data) > 0:
self.external_data[_dn]['time'] = data[col_date].values[-1]
self.external_data[_dn]['value'] = data[col_val].values[-1]
if source == 'USTreasuryYieldCurveRate':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
symbol = self.external_data[_dn]['symbol']
history = self.History(USTreasuryYieldCurveRate, symbol, 1, Resolution.Daily)
history = history.reset_index()
if len(history) > 0:
if col_val in history.columns:
self.external_data[_dn]['time'] = pd.to_datetime(history[col_date], utc=True).iloc[0].replace(tzinfo=None)
self.external_data[_dn]['value'] = history[col_val].values[0]
def Prepare_Model_Data(self):
# self.Log(f"{str(self.Time)}: Preparing Model Data")
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
cols_data = self.model_settings['cols_data']
col_fundamental = self.model_settings['col_fundamental']
col_technical = self.model_settings['col_technical']
start_year = self.model_settings['start_year']
trade_hour = self.model_settings['trade_hour']
scaled_tickers = self.model_settings['scaled_tickers']
prediction_lookforward_days = self.model_settings['prediction_lookforward_days']
col_target_gains = f"gains_N{self.model_settings['prediction_lookforward_days']}D"
inflation_map_dict = self.model_settings['inflation_map_dict']
to_save_data = False
for lstm_ticker in self.general_setting["lstm_tickers"]:
data_df = self.ModelData[lstm_ticker].copy()
has_new_data = False
if len(self.data_list_tickers[lstm_ticker]) > 0:
has_new_data = True
data_df_new = pd.DataFrame(self.data_list_tickers[lstm_ticker]).copy()
if lstm_ticker in scaled_tickers:
data_df_new[col_price] = data_df_new[col_price] / 100
data_df_new[col_price_cur] = data_df_new[col_price_cur] / 100
data_df_new = data_df_new[cols_data]
data_df_new['year'] = data_df_new['datetime'].dt.year
data_df_new['hour'] = data_df_new['datetime'].dt.hour
data_df_new['month'] = data_df_new['datetime'].dt.month
data_df_new['year_month'] = data_df_new['year'].astype(str) + "-" + data_df_new['month'].astype(str).apply(lambda s: s.zfill(2))
data_df = pd.concat([data_df, data_df_new])
if len(data_df) > 0:
if (not self.has_initialized_model_data[lstm_ticker]) or has_new_data:
self.has_initialized_model_data[lstm_ticker] = True
data_df.reset_index(drop=True, inplace=True)
if self.TO_SAVE_DATA:
data_df.drop_duplicates('datetime', keep='last', inplace=True)
else:
data_df.drop_duplicates('datetime', keep='first', inplace=True)
data_df.reset_index(drop=True, inplace=True)
data_df.sort_values('datetime', ascending=True, inplace=True)
data_df.reset_index(drop=True, inplace=True)
for col in col_fundamental + col_technical:
data_df[col] = data_df[col].fillna(method='ffill')
data_df = data_df[data_df['year'] >= start_year]
data_df = data_df[data_df['hour'] == trade_hour]
data_df.reset_index(drop=True, inplace=True)
for col in inflation_map_dict:
col_cpi = inflation_map_dict[col]
data_df[f"{col}_d1"] = (data_df[col_cpi] - data_df[col_cpi].shift(1)) / data_df[col_cpi].shift(1)
data_df[f"{col}_d30"] = (data_df[col_cpi] - data_df[col_cpi].shift(30)) / data_df[col_cpi].shift(30)
data_df[col_target_gains] = data_df[col_price].shift(-prediction_lookforward_days) - data_df[col_price]
self.ModelData[lstm_ticker] = data_df.copy()
self.data_list_tickers[lstm_ticker] = []
to_save_data = True
if to_save_data and self.TO_SAVE_DATA:
self.ObjectStore.SaveBytes(f"MODEL_DATA_{self.model_name}", pickle.dumps(self.ModelData))
self.Log(f"{str(self.Time)}: Model Data Saved")
# self.Log(f"{str(self.Time)}: Model Data Prepared")
def Get_Month_Start_Date(self):
self.month_start_date = self.Time
def Train_Model_Both(self):
model_type = "both"
# self.Log(f"{str(self.Time)}: {model_type}: Training Model")
model_setting = self.model_settings[f"model_settings_{model_type}"]
col_date = self.model_settings['col_date']
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
col_target = self.model_settings['col_target']
prediction_lookforward_days = self.model_settings['prediction_lookforward_days']
col_target_gains = f"gains_N{self.model_settings['prediction_lookforward_days']}D"
use_gru_model = model_setting['use_gru_model']
use_dual_lstm = model_setting['use_dual_lstm']
epochs = model_setting['epochs']
hidden_size = model_setting['hidden_size']
window_size = model_setting['window_size']
thres_multiplier = model_setting['thres_multiplier']
use_early_stop = model_setting['use_early_stop']
learning_rate = model_setting['learning_rate']
batch_size = model_setting['batch_size']
use_weighted_sampler = model_setting['use_weighted_sampler']
volatility_type = model_setting['volatility_type']
valid_lookback_months = model_setting['valid_lookback_months']
train_lookback_months = model_setting['train_lookback_months']
inflation_map_dict = self.model_settings['inflation_map_dict']
for lstm_ticker in self.general_setting["lstm_tickers"]:
if self.month_start_date is None:
continue
else:
month_start_year = self.month_start_date.year
month_start_month = self.month_start_date.month
month_start_day = self.month_start_date.day
model_train_day = month_start_day + self.general_setting["lstm_model_training_displace_days"][lstm_ticker]
to_train = (self.Time.year == month_start_year) and (self.Time.month == month_start_month) and (self.Time.day == model_train_day)
if not to_train:
continue
data_df = self.ModelData[lstm_ticker].copy()
if len(data_df) == 0:
continue
col_feature_both = model_setting["col_feature_dict"][lstm_ticker]
year_month_list = sorted(list(set(data_df['year_month'])))
year_month_vec = np.array(year_month_list)
year_vec = np.array(sorted(list(set(data_df['year']))))
test_year_month = f"{self.Time.year}-{str(self.Time.month).zfill(2)}"
valid_year_month_list = list(year_month_vec[year_month_vec < test_year_month][-valid_lookback_months:])
if len(valid_year_month_list) < valid_lookback_months:
continue
if np.sum(year_month_vec < min(valid_year_month_list)) == 0:
continue
train_year_month_list = list(year_month_vec[year_month_vec < min(valid_year_month_list)][-train_lookback_months:])
if len(train_year_month_list) < train_lookback_months:
continue
data_df_temp = data_df.copy()
if volatility_type == 'thres_v1':
col_target_gains_thres = 0.00200
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
if volatility_type == 'thres_v2':
col_target_gains_thres = 0.00235
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
if volatility_type == 'thres_auto_v1':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
if volatility_type == 'thres_auto_v2':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list+valid_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
self.ModelParams[lstm_ticker][model_type]['col_target_gains_thres'] = col_target_gains_thres
data_df_temp = data_df_temp.dropna()
data_df_temp.reset_index(drop=True, inplace=True)
train_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
valid_df = data_df_temp[data_df_temp['year_month'].isin(valid_year_month_list)].copy()
valid_df_windowed = pd.concat([train_df,valid_df]).copy()
valid_df_windowed = valid_df_windowed.tail(len(valid_df) + window_size-1)
set_seed(100)
(train_loader, val_loader,
_, scaler, weighted_sampler, class_weights) = get_torch_rnn_dataloaders_both(
[col_price] + col_feature_both, col_target, train_df, valid_df_windowed, None, window_size, batch_size,
use_weighted_sampler=use_weighted_sampler,
has_test_data=False,
)
self.Scalers[lstm_ticker][model_type]['both'] = None
self.Scalers[lstm_ticker][model_type]['both'] = scaler
self.Models[lstm_ticker][model_type]['both'] = None
self.Models[lstm_ticker][model_type]['both'] = get_rnn_model_both(
[col_price] + col_feature_both, train_loader, val_loader,
epochs, batch_size, learning_rate, window_size, hidden_size, device,
use_early_stop=use_early_stop, use_weighted_sampler=use_weighted_sampler, class_weights=class_weights,
use_dual_lstm=use_dual_lstm, use_gru_model=use_gru_model,
)
self.Log(f"{str(self.Time)}: {model_type}: {lstm_ticker}: Model Trained")
def Train_Model_Hybrid(self):
model_type = "hybrid"
# self.Log(f"{str(self.Time)}: {model_type}: Training Model")
model_setting = self.model_settings[f"model_settings_{model_type}"]
col_date = self.model_settings['col_date']
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
col_target = self.model_settings['col_target']
prediction_lookforward_days = self.model_settings['prediction_lookforward_days']
col_target_gains = f"gains_N{self.model_settings['prediction_lookforward_days']}D"
use_gru_model = model_setting['use_gru_model']
use_dual_lstm = model_setting['use_dual_lstm']
epochs = model_setting['epochs']
hidden_size = model_setting['hidden_size']
window_size = model_setting['window_size']
thres_multiplier = model_setting['thres_multiplier']
learning_rate = model_setting['learning_rate']
batch_size = model_setting['batch_size']
volatility_type = model_setting['volatility_type']
valid_lookback_months = model_setting['valid_lookback_months']
train_lookback_months = model_setting['train_lookback_months']
inflation_map_dict = self.model_settings['inflation_map_dict']
for lstm_ticker in self.general_setting["lstm_tickers"]:
if self.month_start_date is None:
continue
else:
month_start_year = self.month_start_date.year
month_start_month = self.month_start_date.month
month_start_day = self.month_start_date.day
model_train_day = month_start_day + self.general_setting["lstm_model_training_displace_days"][lstm_ticker]
to_train = (self.Time.year == month_start_year) and (self.Time.month == month_start_month) and (self.Time.day == model_train_day)
if not to_train:
continue
data_df = self.ModelData[lstm_ticker]
if len(data_df) == 0:
continue
col_feature_fundamental = model_setting["col_feature_fundamental_dict"][lstm_ticker]
col_feature_technical = model_setting["col_feature_technical_dict"][lstm_ticker]
year_month_list = sorted(list(set(data_df['year_month'])))
year_month_vec = np.array(year_month_list)
year_vec = np.array(sorted(list(set(data_df['year']))))
test_year_month = f"{self.Time.year}-{str(self.Time.month).zfill(2)}"
valid_year_month_list = list(year_month_vec[year_month_vec < test_year_month][-valid_lookback_months:])
if len(valid_year_month_list) < valid_lookback_months:
continue
if np.sum(year_month_vec < min(valid_year_month_list)) == 0:
continue
train_year_month_list = list(year_month_vec[year_month_vec < min(valid_year_month_list)][-train_lookback_months:])
if len(train_year_month_list) < train_lookback_months:
continue
data_df_temp = data_df.copy()
if volatility_type == 'thres_v1':
col_target_gains_thres = 0.00200
if volatility_type == 'thres_v2':
col_target_gains_thres = 0.00235
if volatility_type == 'thres_auto_v1':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
if volatility_type == 'thres_auto_v2':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list+valid_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
self.ModelParams[lstm_ticker][model_type]['col_target_gains_thres'] = col_target_gains_thres
data_df_temp[col_target] = data_df_temp[col_price].shift(-prediction_lookforward_days)
data_df_temp = data_df_temp.dropna()
data_df_temp.reset_index(drop=True, inplace=True)
train_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
valid_df = data_df_temp[data_df_temp['year_month'].isin(valid_year_month_list)].copy()
set_seed(100)
(train_loader, val_loader, _, scaler) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_fundamental, col_target, train_df, valid_df, None, window_size, batch_size,
has_test_data=False,
)
self.Scalers[lstm_ticker][model_type]['fundamental'] = None
self.Scalers[lstm_ticker][model_type]['fundamental'] = scaler
self.Models[lstm_ticker][model_type]['fundamental'] = None
self.Models[lstm_ticker][model_type]['fundamental'] = get_rnn_model_hybrid(
[col_price] + col_feature_fundamental, train_loader, val_loader,
epochs, learning_rate, hidden_size, device,
use_dual_lstm=use_dual_lstm, use_gru_model=use_gru_model,
)
y_pred_val = get_predictions_hybrid(
val_loader,
self.Models[lstm_ticker][model_type]['fundamental'],
self.Scalers[lstm_ticker][model_type]['fundamental'],
[col_price] + col_feature_fundamental,
device,
)
valid_df['pred_price_fundamental'] = y_pred_val
valid_df['pred_fundamental'] = (valid_df['pred_price_fundamental'] - valid_df[col_price]).apply(get_regression_pred_decision, col_target_gains_thres=col_target_gains_thres)
set_seed(100)
(train_loader, val_loader, _, scaler) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_technical, col_target, train_df, valid_df, None, window_size, batch_size,
has_test_data=False,
)
self.Scalers[lstm_ticker][model_type]['technical'] = scaler
self.Models[lstm_ticker][model_type]['technical'] = get_rnn_model_hybrid(
[col_price] + col_feature_technical, train_loader, val_loader,
epochs, learning_rate, hidden_size, device,
use_dual_lstm=use_dual_lstm, use_gru_model=use_gru_model,
)
y_pred_val = get_predictions_hybrid(
val_loader,
self.Models[lstm_ticker][model_type]['technical'],
self.Scalers[lstm_ticker][model_type]['technical'],
[col_price] + col_feature_technical,
device,
)
valid_df['pred_price_technical'] = y_pred_val
valid_df['pred_technical'] = (valid_df['pred_price_technical'] - valid_df[col_price]).apply(get_regression_pred_decision, col_target_gains_thres=col_target_gains_thres)
fundamental_mse = mean_squared_error(valid_df['pred_price_fundamental'], valid_df[col_target])
technical_mse = mean_squared_error(valid_df['pred_price_technical'], valid_df[col_target])
self.ModelParams[lstm_ticker][model_type]['fundamental_mse'] = fundamental_mse
self.ModelParams[lstm_ticker][model_type]['technical_mse'] = technical_mse
self.Log(f"{str(self.Time)}: {model_type}: {lstm_ticker}: Model Trained")
def OnData(self, data):
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'NasdaqDataLink':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
self.external_data[_dn]['value'] = data[symbol].Value
if source == 'equity':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
if data[symbol] is not None:
self.external_data[_dn]['time'] = data[symbol].Time
self.external_data[_dn]['value'] = data[symbol].Price
FxLstm_SymbolQuantity = {}
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
ticker = self.symbol_ticker_map[symbol]
is_valid_time = self.Time.minute == 0
is_valid_time = is_valid_time and (self.Time.hour in [self.general_setting['FxLstm_prediction_hour']])
if is_valid_time:
if ticker in self.general_setting["lstm_tickers"]:
data_dict = {}
data_dict["datetime"] = self.Time
# data_dict["ticker"] = ticker
# data_dict["symbol"] = symbol.Value
# data_dict["isWarmingUp"] = self.IsWarmingUp
# data_dict["symbolTime"] = str(data[symbol].Time)
data_dict["price"] = np.round(data[symbol].Price, 6)
# Daily Data
_consolidator = symbolData.consolidators["D1"]
data_dict["close_D1"] = _consolidator.close[0]
# External Data
for _dn in self.general_setting["external_data"]:
# data_dict[f"{_dn}_time"] = self.external_data[_dn]['time']
data_dict[_dn] = self.external_data[_dn]['value']
# Technical Features
for _tf in self.general_setting["features"]:
_consolidator = symbolData.consolidators[_tf]
for _in in self.general_setting["features"][_tf]:
_indicator = _consolidator.indicators[_in]
if _in in self.general_setting["features_val_map"]:
for _v in self.general_setting["features_val_map"][
_in
]:
data_dict[f"{_tf}-{_in}-{_v}"] = np.round(
_indicator[_v][0], 5
)
if self.TO_SAVE_DATA:
if not self.IsWarmingUp:
self.data_list_tickers[ticker] += [data_dict]
else:
self.data_list_tickers[ticker] += [data_dict]
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
cols_data = self.model_settings['cols_data']
col_fundamental = self.model_settings['col_fundamental']
col_technical = self.model_settings['col_technical']
start_year = self.model_settings['start_year']
trade_hour = self.model_settings['trade_hour']
col_target = self.model_settings['col_target']
scaled_tickers = self.model_settings['scaled_tickers']
inflation_map_dict = self.model_settings['inflation_map_dict']
max_window_size = self.model_settings['max_window_size']
test_df = pd.DataFrame()
if len(self.data_list_tickers[ticker]) > 0:
data_df_new = pd.DataFrame(self.data_list_tickers[ticker]).copy()
if ticker in scaled_tickers:
data_df_new[col_price] = data_df_new[col_price] / 100
data_df_new[col_price_cur] = data_df_new[col_price_cur] / 100
data_df_new = data_df_new[cols_data]
data_df_new['year'] = data_df_new['datetime'].dt.year
data_df_new['hour'] = data_df_new['datetime'].dt.hour
data_df_new['month'] = data_df_new['datetime'].dt.month
data_df_new['year_month'] = data_df_new['year'].astype(str) + "-" + data_df_new['month'].astype(str).apply(lambda s: s.zfill(2))
data_df = self.ModelData[ticker].copy()
if len(data_df) > 0:
idx = data_df['datetime'] < self.Time
data_df = data_df[idx]
data_df.reset_index(drop=True, inplace=True)
data_df = pd.concat([data_df, data_df_new])
data_df.reset_index(drop=True, inplace=True)
data_df.drop_duplicates('datetime', keep='last', inplace=True)
data_df.reset_index(drop=True, inplace=True)
data_df.sort_values('datetime', ascending=True, inplace=True)
data_df.reset_index(drop=True, inplace=True)
for col in col_fundamental + col_technical:
data_df[col] = data_df[col].fillna(method='ffill')
data_df = data_df[data_df['year'] >= start_year]
data_df = data_df[data_df['hour'] == trade_hour]
data_df.reset_index(drop=True, inplace=True)
for col in inflation_map_dict:
col_cpi = inflation_map_dict[col]
data_df[f"{col}_d1"] = (data_df[col_cpi] - data_df[col_cpi].shift(1)) / data_df[col_cpi].shift(1)
data_df[f"{col}_d30"] = (data_df[col_cpi] - data_df[col_cpi].shift(30)) / data_df[col_cpi].shift(30)
test_df = data_df.tail(max_window_size).copy()
test_df.reset_index(drop=True, inplace=True)
for model_type in self.general_setting["model_types"]:
if len(test_df) == 0:
continue
if model_type == 'both':
if self.Models[ticker][model_type]['both'] is None:
continue
if model_type == 'hybrid':
if self.Models[ticker][model_type]['fundamental'] is None:
continue
if self.Models[ticker][model_type]['technical'] is None:
continue
model_setting = self.model_settings[f"model_settings_{model_type}"]
test_df_windowed = test_df.tail(model_setting['window_size']).copy()
test_df_windowed.reset_index(drop=True, inplace=True)
if len(test_df_windowed) != model_setting['window_size']:
continue
if model_type == 'both':
col_feature_both = model_setting["col_feature_dict"][ticker]
test_df_windowed[col_target] = 1
(_, _, test_loader, _, _, _) = get_torch_rnn_dataloaders_both(
[col_price] + col_feature_both, col_target, None, None, test_df_windowed.copy(),
model_setting['window_size'], model_setting['batch_size'],
use_weighted_sampler=False,
has_test_data=True,
is_training=False,
scaler=self.Scalers[ticker][model_type]['both'],
)
(y_pred_list, y_score_list) = get_predictions_both(test_loader, self.Models[ticker][model_type]['both'], device)
y_pred = y_pred_list[-1]
if model_type == 'hybrid':
col_feature_fundamental = model_setting["col_feature_fundamental_dict"][ticker]
col_feature_technical = model_setting["col_feature_technical_dict"][ticker]
test_df_windowed[col_target] = 1
ref_price = test_df_windowed[col_price].values[-1]
col_target_gains_thres = self.ModelParams[ticker][model_type]['col_target_gains_thres']
fundamental_mse = self.ModelParams[ticker][model_type]['fundamental_mse']
technical_mse = self.ModelParams[ticker][model_type]['technical_mse']
(_, _, test_loader, _) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_fundamental, col_target, None, None, test_df_windowed.copy(),
model_setting['window_size'], model_setting['batch_size'],
has_test_data=True,
is_training=False,
scaler=self.Scalers[ticker][model_type]['fundamental'],
)
y_pred_val = get_predictions_hybrid(
test_loader,
self.Models[ticker][model_type]['fundamental'],
self.Scalers[ticker][model_type]['fundamental'],
[col_price] + col_feature_fundamental,
device,
)
pred_price_fundamental = y_pred_val[-1]
pred_fundamental = get_regression_pred_decision(pred_price_fundamental - ref_price, col_target_gains_thres)
(_, _, test_loader, _) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_technical, col_target, None, None, test_df_windowed.copy(),
model_setting['window_size'], model_setting['batch_size'],
has_test_data=True,
is_training=False,
scaler=self.Scalers[ticker][model_type]['technical'],
)
y_pred_val = get_predictions_hybrid(
test_loader,
self.Models[ticker][model_type]['technical'],
self.Scalers[ticker][model_type]['technical'],
[col_price] + col_feature_technical,
device,
)
pred_price_technical = y_pred_val[-1]
pred_technical = get_regression_pred_decision(pred_price_technical - ref_price, col_target_gains_thres)
y_pred = get_prediction_hybrid_regression(pred_fundamental, pred_technical, fundamental_mse, technical_mse)
for _signal in self.general_setting["signals"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
pred_type = self.signal_settings[_signal]['pred_type']
lstm_ticker = self.signal_settings[_signal]['lstm_ticker']
if (pred_type == model_type) and (lstm_ticker == ticker):
self.prediction_dict[_signal][ticker] = y_pred
self.Signal[symbol][_signal].update_prediction_direction(self.prediction_dict[_signal][ticker])
symbolQuantity = 0
for _signal in self.general_setting["signals"]:
if ticker in self.signal_settings[_signal]['valid_tickers']:
to_exit = self.Signal[symbol][_signal].check_exit(symbolData, data[symbol].Price, data[symbol].Time)
if to_exit:
self.Signal[symbol][_signal].update_exit()
has_enter = self.Signal[symbol][_signal].enter(symbolData, data[symbol].Price, data[symbol].Time)
quantity = self.Signal[symbol][_signal].quantity * self.Signal[symbol][_signal].allocation_multiplier
quantity = int(np.ceil(quantity))
symbolQuantity += quantity
FxLstm_SymbolQuantity[symbol] = symbolQuantity
## Aggregate symbol quantities across strategies
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
self.Counter[symbol]["counter"] += 1
symbolQuantity = 0
if symbol in FxLstm_SymbolQuantity:
symbolQuantity += FxLstm_SymbolQuantity[symbol]
if not self.IsWarmingUp:
# In case orders takes longer than 1 bar to be filled. Only send market orders every 3 minutes
if (self.Counter[symbol]["counter"] - self.Counter[symbol]["last_order_counter"]) >= self.general_setting["order_counter_diff"]:
if (symbolQuantity - self.Portfolio[symbol].Quantity) != 0:
self.MarketOrder(symbol, symbolQuantity - self.Portfolio[symbol].Quantity)
self.Counter[symbol]["last_order_counter"] = self.Counter[symbol]["counter"]
## Version
# Forex LSTM V 1.0
from AlgorithmImports import *
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from collections import deque
import pickle
from config import (
general_setting,
consolidator_settings,
indicator_settings,
signal_settings,
)
from data_classes import (
SymbolData,
MarketHours,
)
from signal_classes import (
FxLstmSignal,
)
from QuantConnect.DataSource import *
class FxLstmAlgo(QCAlgorithm):
def Initialize(self):
self.SetTimeZone(TimeZones.Johannesburg)
self.SetStartDate(2013, 1, 1)
self.SetCash(10000)
self.SetWarmUp(int(12 * 20 * 24 * 60), Resolution.Minute)
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
self.to_plot = False
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.signal_settings = signal_settings
##-##
self.ref_ticker = "EURUSD"
# self.ref_ticker = "USDJPY"
# self.ref_ticker = "USDCHF"
# self.ref_ticker = "GBPUSD"
# self.ref_ticker = "AUDUSD"
# self.ref_ticker = "USDCAD"
##-##
# Data Initialization
self.Data = {}
self.Counter = {}
self.SymbolMarketHours = {}
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.data_list = []
self.output_data_dict = {}
for ticker in self.general_setting["tickers"]:
if general_setting["tickers"][ticker]["type"] == "equity":
symbol = self.AddEquity(
ticker,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.Raw,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "forex":
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "cfd":
symbol = self.AddCfd(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.Data[symbol] = SymbolData(
self,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
)
self.Counter[symbol] = {}
self.Counter[symbol]["counter"] = 0
self.Counter[symbol]["last_order_counter"] = 0
self.SymbolMarketHours[symbol] = MarketHours(self, symbol)
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
self.Schedule.On(
self.DateRules.MonthEnd(self.ref_ticker),
self.TimeRules.BeforeMarketClose(self.ref_ticker, 0),
self.SaveData,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker),
self.SOD,
)
self.external_data = {}
for _dn in self.general_setting["external_data"]:
self.external_data[_dn] = {}
self.external_data[_dn]['time'] = None
self.external_data[_dn]['value'] = None
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
link = self.general_setting["external_data"][_dn]['link']
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
data = self.Download(link)
rows = []
for row in data.split('\n'):
rows.append(row.replace('\r','').lower().split(','))
data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
data_df[col_val] = data_df[col_val].astype(float)
self.external_data[_dn]['data'] = data_df.copy()
if source == 'NasdaqDataLink':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(NasdaqDataLink, ref, Resolution.Daily).Symbol
if source == 'equity':
ticker = self.general_setting["external_data"][_dn]['ticker']
self.external_data[_dn]['symbol'] = self.AddEquity(ticker, Resolution.Daily).Symbol
if source == 'USTreasuryYieldCurveRate':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(USTreasuryYieldCurveRate, ref).Symbol
def SOD(self):
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
lag_days = self.general_setting["external_data"][_dn]['lag_days']
data = self.external_data[_dn]['data'][self.external_data[_dn]['data'][col_date] < (self.Time - timedelta(days=lag_days))]
if len(data) > 0:
self.external_data[_dn]['time'] = data[col_date].values[-1]
self.external_data[_dn]['value'] = data[col_val].values[-1]
if source == 'USTreasuryYieldCurveRate':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
symbol = self.external_data[_dn]['symbol']
history = self.History(USTreasuryYieldCurveRate, symbol, 1, Resolution.Daily)
history = history.reset_index()
if len(history) > 0:
if col_val in history.columns:
self.external_data[_dn]['time'] = pd.to_datetime(history[col_date], utc=True).iloc[0].replace(tzinfo=None)
self.external_data[_dn]['value'] = history[col_val].values[0]
def SaveData(self):
self.output_data_dict["data"] = self.data_list
self.output_data_dict["version"] = "1"
self.ObjectStore.SaveBytes(f"FUNDAMENTAL_DATA_{self.ref_ticker}", pickle.dumps(self.output_data_dict))
# self.ObjectStore.SaveBytes(f"TECHNICAL_DATA_{self.ref_ticker}", pickle.dumps(self.output_data_dict))
def OnData(self, data):
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'NasdaqDataLink':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
self.external_data[_dn]['value'] = data[symbol].Value
if source == 'equity':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
if data[symbol] is not None:
self.external_data[_dn]['time'] = data[symbol].Time
self.external_data[_dn]['value'] = data[symbol].Price
FxLstm_SymbolQuantity = {}
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
ticker = self.symbol_ticker_map[symbol]
is_valid_time = self.Time.minute == 0
# is_valid_time = is_valid_time and (self.Time.hour in [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23])
is_valid_time = is_valid_time and (self.Time.hour in [1])
if is_valid_time:
if ticker in self.general_setting["lstm_tickers"]:
data_dict = {}
data_dict["datetime"] = self.Time
# data_dict["ticker"] = ticker
# data_dict["symbol"] = symbol.Value
# data_dict["isWarmingUp"] = self.IsWarmingUp
# data_dict["symbolTime"] = str(data[symbol].Time)
# data_dict["price"] = np.round(data[symbol].Price, 6)
# Daily Data
_consolidator = symbolData.consolidators["D1"]
# data_dict["close_D1"] = _consolidator.close[0]
# External Data
for _dn in self.general_setting["external_data"]:
# data_dict[f"{_dn}_time"] = self.external_data[_dn]['time']
data_dict[_dn] = self.external_data[_dn]['value']
# Technical Features
# for _tf in self.general_setting["features"]:
# _consolidator = symbolData.consolidators[_tf]
# for _in in self.general_setting["features"][_tf]:
# _indicator = _consolidator.indicators[_in]
# if _in in self.general_setting["features_val_map"]:
# for _v in self.general_setting["features_val_map"][
# _in
# ]:
# data_dict[f"{_tf}-{_in}-{_v}"] = np.round(
# _indicator[_v][0], 5
# )
if not self.IsWarmingUp:
self.data_list += [data_dict]
## Version
# Forex LSTM V 1.0
from AlgorithmImports import *
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from collections import deque
import pickle
from config import (
general_setting,
consolidator_settings,
indicator_settings,
signal_settings,
)
from data_classes import (
SymbolData,
MarketHours,
)
from signal_classes import (
FxLstmSignal,
)
from QuantConnect.DataSource import *
class FxLstmAlgo(QCAlgorithm):
def Initialize(self):
self.SetTimeZone(TimeZones.Johannesburg)
self.SetStartDate(2013, 1, 1)
self.SetCash(10000)
self.SetWarmUp(int(12 * 20 * 24 * 60), Resolution.Minute)
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
self.to_plot = False
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.signal_settings = signal_settings
##-##
self.ref_ticker = "EURUSD"
# self.ref_ticker = "USDJPY"
# self.ref_ticker = "USDCHF"
# self.ref_ticker = "GBPUSD"
# self.ref_ticker = "AUDUSD"
# self.ref_ticker = "USDCAD"
##-##
# Data Initialization
self.Data = {}
self.Counter = {}
self.SymbolMarketHours = {}
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.data_list = []
self.output_data_dict = {}
for ticker in self.general_setting["tickers"]:
if general_setting["tickers"][ticker]["type"] == "equity":
symbol = self.AddEquity(
ticker,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.Raw,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "forex":
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "cfd":
symbol = self.AddCfd(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.Data[symbol] = SymbolData(
self,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
)
self.Counter[symbol] = {}
self.Counter[symbol]["counter"] = 0
self.Counter[symbol]["last_order_counter"] = 0
self.SymbolMarketHours[symbol] = MarketHours(self, symbol)
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
self.Schedule.On(
self.DateRules.MonthEnd(self.ref_ticker),
self.TimeRules.BeforeMarketClose(self.ref_ticker, 0),
self.SaveData,
)
self.Schedule.On(
self.DateRules.EveryDay(self.ref_ticker),
self.TimeRules.AfterMarketOpen(self.ref_ticker),
self.SOD,
)
self.external_data = {}
for _dn in self.general_setting["external_data"]:
self.external_data[_dn] = {}
self.external_data[_dn]['time'] = None
self.external_data[_dn]['value'] = None
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
link = self.general_setting["external_data"][_dn]['link']
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
data = self.Download(link)
rows = []
for row in data.split('\n'):
rows.append(row.replace('\r','').lower().split(','))
data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
data_df[col_val] = data_df[col_val].astype(float)
self.external_data[_dn]['data'] = data_df.copy()
if source == 'NasdaqDataLink':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(NasdaqDataLink, ref, Resolution.Daily).Symbol
if source == 'equity':
ticker = self.general_setting["external_data"][_dn]['ticker']
self.external_data[_dn]['symbol'] = self.AddEquity(ticker, Resolution.Daily).Symbol
if source == 'USTreasuryYieldCurveRate':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(USTreasuryYieldCurveRate, ref).Symbol
def SOD(self):
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
lag_days = self.general_setting["external_data"][_dn]['lag_days']
data = self.external_data[_dn]['data'][self.external_data[_dn]['data'][col_date] < (self.Time - timedelta(days=lag_days))]
if len(data) > 0:
self.external_data[_dn]['time'] = data[col_date].values[-1]
self.external_data[_dn]['value'] = data[col_val].values[-1]
if source == 'USTreasuryYieldCurveRate':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
symbol = self.external_data[_dn]['symbol']
history = self.History(USTreasuryYieldCurveRate, symbol, 1, Resolution.Daily)
history = history.reset_index()
if len(history) > 0:
if col_val in history.columns:
self.external_data[_dn]['time'] = pd.to_datetime(history[col_date], utc=True).iloc[0].replace(tzinfo=None)
self.external_data[_dn]['value'] = history[col_val].values[0]
def SaveData(self):
self.output_data_dict["data"] = self.data_list
self.output_data_dict["version"] = "1"
# self.ObjectStore.SaveBytes(f"FUNDAMENTAL_DATA_{self.ref_ticker}", pickle.dumps(self.output_data_dict))
self.ObjectStore.SaveBytes(f"TECHNICAL_DATA_{self.ref_ticker}", pickle.dumps(self.output_data_dict))
def OnData(self, data):
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'NasdaqDataLink':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
self.external_data[_dn]['value'] = data[symbol].Value
if source == 'equity':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
if data[symbol] is not None:
self.external_data[_dn]['time'] = data[symbol].Time
self.external_data[_dn]['value'] = data[symbol].Price
FxLstm_SymbolQuantity = {}
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
ticker = self.symbol_ticker_map[symbol]
is_valid_time = self.Time.minute == 0
# is_valid_time = is_valid_time and (self.Time.hour in [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23])
is_valid_time = is_valid_time and (self.Time.hour in [1])
if is_valid_time:
if ticker in self.general_setting["lstm_tickers"]:
data_dict = {}
data_dict["datetime"] = self.Time
# data_dict["ticker"] = ticker
# data_dict["symbol"] = symbol.Value
# data_dict["isWarmingUp"] = self.IsWarmingUp
# data_dict["symbolTime"] = str(data[symbol].Time)
data_dict["price"] = np.round(data[symbol].Price, 6)
# Daily Data
_consolidator = symbolData.consolidators["D1"]
data_dict["close_D1"] = _consolidator.close[0]
# External Data
for _dn in self.general_setting["external_data"]:
data_dict[f"{_dn}_time"] = self.external_data[_dn]['time']
data_dict[_dn] = self.external_data[_dn]['value']
# Technical Features
for _tf in self.general_setting["features"]:
_consolidator = symbolData.consolidators[_tf]
for _in in self.general_setting["features"][_tf]:
_indicator = _consolidator.indicators[_in]
if _in in self.general_setting["features_val_map"]:
for _v in self.general_setting["features_val_map"][
_in
]:
data_dict[f"{_tf}-{_in}-{_v}"] = np.round(
_indicator[_v][0], 5
)
if not self.IsWarmingUp:
self.data_list += [data_dict]
# region imports
from AlgorithmImports import *
# endregion
from datetime import datetime
import numpy as np
TRADE_FUTURES = False
EXIT_2023 = True
class FuturesBasis(QCAlgorithm):
def Initialize(self):
##-##
# if EXIT_2023:
# self.SetStartDate(datetime(2020, 1, 1))
# self.SetEndDate(datetime(2023, 5, 30))
# else:
# self.SetStartDate(datetime(2023, 1, 1))
self.SetStartDate(datetime(2020, 1, 1))
self.SetEndDate(datetime(2023, 5, 30))
##-##
self.SetCash(100000)
if TRADE_FUTURES:
self.SetBrokerageModel(BrokerageName.InteractiveBrokersBrokerage, AccountType.Margin)
else:
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
##-##
# self.ticker_list = ['EURUSD']
# self.ticker_list = ['GBPUSD']
# self.ticker_list = ['AUDUSD']
self.ticker_list = ['USDJPY']
self.future_units_dict = {
"EURUSD": 12500,
"GBPUSD": 6250,
"AUDUSD": 10000,
"USDJPY": -50000,
}
##-##
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.Futures = {}
self.FuturesTracker = {}
self.FuturesSymbol = {}
self.FuturesRefSymbol = {}
for ticker in self.ticker_list:
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
if ticker == "EURUSD":
self.Futures[ticker] = self.AddFuture(
# Futures.Currencies.EUR,
# Futures.Currencies.EuroFXEmini,
Futures.Currencies.MicroEUR,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.BackwardsRatio,
dataMappingMode=DataMappingMode.FirstDayMonth
)
if ticker == "GBPUSD":
self.Futures[ticker] = self.AddFuture(
Futures.Currencies.MicroGBP,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.BackwardsRatio,
dataMappingMode=DataMappingMode.FirstDayMonth
)
if ticker == "AUDUSD":
self.Futures[ticker] = self.AddFuture(
Futures.Currencies.MicroAUD,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.BackwardsRatio,
dataMappingMode=DataMappingMode.FirstDayMonth
)
if ticker == "USDJPY":
self.Futures[ticker] = self.AddFuture(
## Futures.Currencies.MicroJPY,
## Futures.Currencies.MicroUSDJPY,
Futures.Currencies.JapaneseYenEmini,
# Futures.Currencies.JPY,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.BackwardsRatio,
dataMappingMode=DataMappingMode.FirstDayMonth
)
self.Futures[ticker].SetFilter(0, 182)
self.FuturesSymbol[ticker] = self.Futures[ticker].Symbol
self.FuturesRefSymbol[ticker] = None
self.FuturesTracker[ticker] = None
self.traded = False
def OnData(self, data):
# to_continue = False
# for ticker in self.ticker_list:
# if not data.ContainsKey(self.ticker_symbol_map[ticker]):
# to_continue = True
# else:
# if data[self.ticker_symbol_map[ticker]] is None:
# to_continue = True
# if not data.ContainsKey(self.FuturesSymbol[ticker]):
# to_continue = True
# else:
# if data[self.FuturesSymbol[ticker]] is None:
# to_continue = True
# if to_continue:
# return
for ticker in self.ticker_list:
# self.Log(f"symbol: {self.FuturesSymbol[ticker]}")
##_##
# if self.FuturesTracker[ticker] is not None:
# quantity = self.Portfolio[self.FuturesTracker[ticker]].Quantity
# if quantity == 0:
# self.FuturesTracker[ticker] = None
##_##
chain = data.FuturesChains.get(self.FuturesSymbol[ticker])
if chain:
# Select the contract with the greatest open interest
ref_contract = sorted(chain, key=lambda contract: contract.OpenInterest, reverse=True)[0]
self.FuturesRefSymbol[ticker] = ref_contract.Symbol
if data.SymbolChangedEvents.ContainsKey(self.FuturesSymbol[ticker]):
# self.Log("event change")
changed_event = data.SymbolChangedEvents[self.FuturesSymbol[ticker]]
old_symbol = changed_event.OldSymbol
new_symbol = changed_event.NewSymbol
if self.FuturesRefSymbol[ticker] != old_symbol:
rollover_symbol = self.FuturesRefSymbol[ticker]
else:
rollover_symbol = new_symbol
# tag = f"Rollover - Symbol changed at {self.Time}: {old_symbol} -> {new_symbol}"
# self.Log(tag)
if self.FuturesTracker[ticker] is not None:
if old_symbol == self.FuturesTracker[ticker]:
if TRADE_FUTURES:
quantity = self.Portfolio[old_symbol].Quantity
if quantity != 0:
self.MarketOrder(old_symbol, -quantity)
self.MarketOrder(rollover_symbol, quantity)
else:
quantity = self.Portfolio[self.ticker_symbol_map[ticker]].Quantity
if quantity != 0:
self.MarketOrder(self.ticker_symbol_map[ticker], -quantity)
self.MarketOrder(self.ticker_symbol_map[ticker], quantity)
self.FuturesTracker[ticker] = rollover_symbol
##-##
if not self.traded and (self.FuturesRefSymbol[ticker] is not None):
# if (self.FuturesTracker[ticker] is None) and (self.FuturesRefSymbol[ticker] is not None) and ((not EXIT_2023) or (EXIT_2023 and (self.Time.year < 2023))):
##-##
self.traded = True
self.FuturesTracker[ticker] = self.FuturesRefSymbol[ticker]
if TRADE_FUTURES:
self.MarketOrder(self.FuturesTracker[ticker], 1)
else:
self.MarketOrder(self.ticker_symbol_map[ticker], self.future_units_dict[ticker])
if self.FuturesTracker[ticker] is not None:
##-##
# quantity = self.Portfolio[self.FuturesTracker[ticker]].Quantity
# if quantity == 0:
# self.FuturesTracker[ticker] = None
# else:
# if self.Time.year >= 2023:
# if TRADE_FUTURES:
# self.MarketOrder(self.FuturesTracker[ticker], -quantity)
# self.FuturesTracker[ticker] = None
# if EXIT_2023 and (self.Time.year >= 2023):
# if TRADE_FUTURES:
# quantity = self.Portfolio[self.FuturesTracker[ticker]].Quantity
# self.MarketOrder(self.FuturesTracker[ticker], -quantity)
# else:
# quantity = self.Portfolio[self.ticker_symbol_map[ticker]].Quantity
# self.MarketOrder(self.ticker_symbol_map[ticker], -quantity)
# self.FuturesTracker[ticker] = None
if self.Time.year >= 2023:
if TRADE_FUTURES:
quantity = self.Portfolio[self.FuturesTracker[ticker]].Quantity
self.MarketOrder(self.FuturesTracker[ticker], -quantity)
else:
quantity = self.Portfolio[self.ticker_symbol_map[ticker]].Quantity
self.MarketOrder(self.ticker_symbol_map[ticker], -quantity)
self.FuturesTracker[ticker] = None
##-##Notebook too long to render.
#region imports
from AlgorithmImports import *
#endregion
## General Settings
general_setting = {
"tickers": {
"USDJPY": {"type": "forex"},
},
"model_name": "ForexLSTM_OANDA_V1_02",
"consolidator_timeframes": ["D1", "W1"],
"order_counter_diff": 3,
"model_types": ["both","hybrid"],
"lstm_tickers": ['USDJPY'],
"lstm_model_training_displace_days": {
'USDJPY': 0,
},
"signals": [
"FxLstm_Hybrid_USDJPY_Trail",
],
"FxLstm_prediction_hour": 1,
"external_data": {
# SP500
'spy': {
'source': 'equity',
'ticker': 'SPY',
},
# Global X DAX Germany ETF
'dax': {
'source': 'equity',
'ticker': 'DAX',
},
# US Treasury
'us_treasury': {
'source': 'USTreasuryYieldCurveRate',
'ref': 'USTYCR',
'col_date': 'time',
'col_val': 'onemonth',
},
# # Consumer Price Index for Inflation Rate
# # https://data.nasdaq.com/data/RATEINF-inflation-rates
# 'cpi_usa': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_USA",
# },
# 'cpi_eur': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_EUR",
# },
# 'cpi_deu': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_DEU",
# },
# 'cpi_gbr': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_GBR",
# },
# 'cpi_chf': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_CHE",
# },
# 'cpi_jpn': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_JPN",
# },
# 'cpi_can': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_CAN",
# },
# 'cpi_aus': {
# 'source': 'NasdaqDataLink',
# 'ref': "RATEINF/CPI_AUS",
# },
# Consumer Price Index: All Items: Total for United States
# https://fred.stlouisfed.org/series/USACPALTT01CTGYM
'cpi_usa': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_usa/data.csv",
'col_date': 'date',
'col_val': 'USACPALTT01CTGYM'.lower(),
'lag_days': 10,
},
# Consumer Price Index: All Items: Total: Total for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/EA19CPALTT01GYM
'cpi_eur': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_eur/data.csv",
'col_date': 'date',
'col_val': 'EA19CPALTT01GYM'.lower(),
'lag_days': 10,
},
# Consumer Price Index: All Items: Total for Germany
# https://fred.stlouisfed.org/series/DEUCPALTT01CTGYM
'cpi_deu': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_deu/data.csv",
'col_date': 'date',
'col_val': 'DEUCPALTT01CTGYM'.lower(),
'lag_days': 10,
},
# Consumer Price Index: All Items: Total for United Kingdom
# https://fred.stlouisfed.org/series/GBRCPALTT01CTGYM
'cpi_gbr': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_gbr/data.csv",
'col_date': 'date',
'col_val': 'GBRCPALTT01CTGYM'.lower(),
'lag_days': 10,
},
# Consumer Price Index: All Items: Total for Switzerland
# https://fred.stlouisfed.org/series/CHECPALTT01CTGYM
'cpi_chf': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_chf/data.csv",
'col_date': 'date',
'col_val': 'CHECPALTT01CTGYM'.lower(),
'lag_days': 10,
},
# Consumer Price Index: All Items: Total for Japan
# https://fred.stlouisfed.org/series/CPALTT01JPM659N
'cpi_jpn': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_jpn/data.csv",
'col_date': 'date',
'col_val': 'CPALTT01JPM659N'.lower(),
'lag_days': 10,
},
# # Consumer Price Index: All Items: Total for Canada
# # https://fred.stlouisfed.org/series/CANCPALTT01CTGYM
'cpi_can': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_can/data.csv",
'col_date': 'date',
'col_val': 'CANCPALTT01CTGYM'.lower(),
'lag_days': 10,
},
# Consumer Price Index: All Items: Total for Australia
# https://fred.stlouisfed.org/series/CPALTT01AUQ659N
'cpi_aus': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/cpi_aus/data.csv",
'col_date': 'date',
'col_val': 'CPALTT01AUQ659N'.lower(),
'lag_days': 10,
},
# Federal Funds Effective Rate (DFF)
# https://fred.stlouisfed.org/series/DFF
'dff': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/dff/data.csv",
'col_date': 'date',
'col_val': 'dff',
'lag_days': 1,
},
# Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/IRLTLT01EZM156N
'rate_eur_lt_gov': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_eur_lt_gov/data.csv",
'col_date': 'date',
'col_val': 'IRLTLT01EZM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for the Euro Area (19 Countries)
# https://fred.stlouisfed.org/series/IR3TIB01EZM156N
'rate_eur_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_eur_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01EZM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: Long-Term Government Bond Yields: 10-Year: Main (Including Benchmark) for Germany
# https://fred.stlouisfed.org/series/IRLTLT01DEM156N
'rate_deu_lt_gov': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_deu_lt_gov/data.csv",
'col_date': 'date',
'col_val': 'IRLTLT01DEM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Germany
# https://fred.stlouisfed.org/series/IR3TIB01DEM156N
'rate_deu_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_deu_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01DEM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Switzerland
# https://fred.stlouisfed.org/series/IR3TIB01CHM156N
'rate_chf_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_chf_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01CHM156N'.lower(),
'lag_days': 10,
},
# 3-Month or 90-day Rates and Yields: Interbank Rates for Japan
# https://fred.stlouisfed.org/series/IR3TIB01JPM156N
'rate_jpn_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_jpn_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01JPM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Australia
# https://fred.stlouisfed.org/series/IR3TIB01AUM156N
'rate_aus_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_aus_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01AUM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for Canada
# https://fred.stlouisfed.org/series/IR3TIB01CAM156N
'rate_cnd_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_cnd_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01CAM156N'.lower(),
'lag_days': 10,
},
# Interest Rates: 3-Month or 90-Day Rates and Yields: Interbank Rates: Total for United Kingdom
# https://fred.stlouisfed.org/series/IR3TIB01GBM156N
'rate_gbp_3m_bank': {
'source': 'gsheet',
'link': "https://raw.githubusercontent.com/deerfieldgreen/FRED_data/main/data/rate_gbp_3m_bank/data.csv",
'col_date': 'date',
'col_val': 'IR3TIB01GBM156N'.lower(),
'lag_days': 10,
},
},
"features": {
"D1": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20","PSAR","AO","ROC10","TRIX10",
],
"W1": [],
},
"features_val_map": {
"SMA10": ["val"],
"MACD": ["macd", "macdsignal", "macdhist"],
"ROC2": ["val"],
"MOM4": ["val"],
"RSI10": ["val"],
"BB20": ["upper","lower","mid"],
"CCI20": ["val"],
"ULTOSC": ["val"],
"CHOP": ["val"],
"DX14": ["val"],
"PHASE": ["val"],
"CRSI": ["val"],
"PSAR": ["val"],
"AO": ["val"],
"TRIX10": ["val"],
"ROC10": ["val"],
},
}
## Consolidator Settings
consolidator_settings = {
"D1": {
"timeframe_minutes": 24 * 60,
"consolidation_type": "quote",
"indicators": [
"SMA10","MACD","ROC2","MOM4","RSI10","BB20","CCI20",
"ATR10","ATR14","ATR21",
"PSAR",
"SMA100",
"SMA200",
"SMA21" ,
"AO",
"ROC10",
"ROC2",
"TRIX10",
"SMA50",
],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
"W1": {
"timeframe_minutes": 7 * 24 * 60,
"consolidation_type": "quote",
"indicators": [],
"window": 5,
"window_multiplier_dict": {
"forex": 1,
},
},
}
## Indicators Settings
indicator_settings = {
"SMA5": {
"type": "SMA",
"lookback": 5,
"field": "Close",
"window": 3,
},
"SMA10": {
"type": "SMA",
"lookback": 10,
"field": "Close",
"window": 3,
},
"SMA20": {
"type": "SMA",
"lookback": 20,
"field": "Close",
"window": 3,
},
"SMA21": {
"type": "SMA",
"lookback": 21,
"field": "Close",
"window": 3,
},
"SMA50": {
"type": "SMA",
"lookback": 50,
"field": "Close",
"window": 3,
},
"SMA100": {
"type": "SMA",
"lookback": 100,
"field": "Close",
"window": 3,
},
"SMA200": {
"type": "SMA",
"lookback": 200,
"field": "Close",
"window": 3,
},
"MACD": {
"type": "MACD",
"window": 3,
},
"ROC2": {
"type": "ROC",
"lookback": 2,
"field": "Close",
"window": 3,
},
"ROC10": {
"type": "ROC",
"lookback": 10,
"field": "Close",
"window": 3,
},
"MOM4": {
"type": "MOM",
"lookback": 2,
"field": "Close",
"window": 3,
},
"RSI10": {
"type": "RSI",
"lookback": 10,
"ma_type": "Simple",
"field": "Close",
"window": 3,
},
"BB20": {
"type": "BOLL",
"lookback": 20,
"ma_type": "Simple",
"std": 2,
"field": "Close",
"window": 3,
},
"CCI20": {
"type": "CCI",
"lookback": 20,
"field": "Close",
"window": 3,
},
"ATR10": {
"type": "ATR",
"lookback": 10,
"field": "Close",
"window": 3,
},
"ATR14": {
"type": "ATR",
"lookback": 14,
"field": "Close",
"window": 3,
},
"ATR21": {
"type": "ATR",
"lookback": 21,
"field": "Close",
"window": 3,
},
"ULTOSC": {
"type": "ULTOSC",
"window": 3,
},
"CHOP": {
"type": "CHOP",
"lookback": 52,
"window": 3,
},
"DX14": {
"type": "DX",
"lookback": 14,
"window": 3,
},
"PHASE": {
"type": "PHASE",
"lookback": 15,
"window": 3,
},
"CRSI": {
"type": "CRSI",
"rsi_len": 15,
"rsi_field": "Close",
"rsi_window": 21,
"window": 3,
},
"PSAR": {
"type": "PSAR",
"window": 3,
},
"AO": {
"type": "AO",
"field": "Close",
"window": 3,
},
"TRIX10": {
"type": "TRIX",
"lookback": 10,
"field": "Close",
"window": 3,
},
}
signal_settings = {
"FxLstm_Hybrid_USDJPY_Trail": {
"lstm_ticker": "USDJPY",
"valid_tickers": ["USDJPY"],
"active_timeframe": "D1",
"prediction_direction_map_dict": {
0: -1,
1: 0,
2: 1,
},
"pred_type": 'hybrid',
"exit_wait_period": 0,
"risk_pct": 0.08,# 0.02
"enter_long_trades": True,
"enter_short_trades": True,
"use_sma_filter": True,
"sma_filter_lookback": 10,
"use_ao_filter": True,
"use_roc_filter": True,
"roc_filter_lookback": 2,
"use_trix_filter": True,
"trix_filter_lookback": 10,
"use_macd_filter": True,
"atrLength": 10,
"longStopMultiplier": 0.5,
"shortStopMultiplier": 0.2, # 0.2
"longRiskRewardMultiplier": 3,
"shortRiskRewardMultiplier": 3,
"useTralingStop": False,
"trailStopSize": 3.0, # 5.0
"use_movement_thres_for_stops": False,
"use_trix_for_stops": False,
"use_ao_for_stops": False,
"use_macd_for_stops": False,
"use_roc_for_stops": False,
"use_sma_for_stops": False,
"movement_thres": 0.002,
"use_prediction_direction_to_exit": False,
},
}
model_settings = {
"col_date": ['datetime'],
"col_price": 'close_D1',
"col_price_cur": 'price',
"col_target": 'target',
"start_year": 2017,
"trade_hour": 1,
"prediction_lookforward_days": 1,
"max_window_size": 100,
"scaled_tickers": ["USDJPY"],
"inflation_map_dict": {
'inflation_usa': 'cpi_usa',
'inflation_eur': 'cpi_eur',
'inflation_deu': 'cpi_deu',
'inflation_gbr': 'cpi_gbr',
'inflation_chf': 'cpi_chf',
'inflation_jpn': 'cpi_jpn',
'inflation_aus': 'cpi_aus',
},
"cols_data": [
'datetime', 'close_D1', 'price',
'spy', 'dax',
'dff',
'cpi_usa', 'cpi_eur', 'cpi_deu', 'cpi_gbr','cpi_chf','cpi_jpn','cpi_aus',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
'rate_chf_3m_bank',
'rate_jpn_3m_bank',
'rate_aus_3m_bank',
'rate_cnd_3m_bank',
'rate_gbp_3m_bank',
'D1-SMA10-val', 'D1-MACD-macd', 'D1-MACD-macdsignal', 'D1-MACD-macdhist', 'D1-ROC2-val',
'D1-MOM4-val', 'D1-RSI10-val', 'D1-BB20-upper', 'D1-BB20-mid', 'D1-BB20-lower', 'D1-CCI20-val',
'D1-PSAR-val',
],
'col_fundamental': [
'spy','dax',
'dff',
'cpi_usa','cpi_eur','cpi_deu','cpi_gbr','cpi_chf','cpi_jpn','cpi_aus',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
'rate_chf_3m_bank',
'rate_jpn_3m_bank',
'rate_aus_3m_bank',
'rate_cnd_3m_bank',
'rate_gbp_3m_bank',
],
'col_technical': [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
'D1-PSAR-val',
],
"model_settings_both": {
"use_gru_model": True,
"use_dual_lstm": False,
"epochs": 1,
"hidden_size": 50,
"window_size": 5,
"thres_multiplier": 3,
"use_early_stop": False,
"learning_rate": 0.0005,
"batch_size": 8,
"use_weighted_sampler": False,
"volatility_type": 'thres_auto_v1',
"valid_lookback_months": 12,
"train_lookback_months": 48,
"col_feature_dict": {
"USDJPY": [
'dff',
'inflation_usa_d30',
'inflation_eur_d30',
'inflation_deu_d30',
'rate_eur_3m_bank',
'rate_deu_3m_bank',
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
'D1-PSAR-val',
],
},
},
"model_settings_hybrid": {
"use_gru_model": True,
"use_dual_lstm": False,
"epochs": 1,
"hidden_size": 50,
"window_size": 20,
"thres_multiplier": 3,
"learning_rate": 0.001,
"batch_size": 8,
"use_weighted_sampler": False,
"volatility_type": 'thres_auto_v1',
"valid_lookback_months": 12,
"train_lookback_months": 48,
"col_feature_fundamental_dict": {
"USDJPY": [
'dff',
# 'inflation_usa_d1',
# 'inflation_deu_d1',
'rate_jpn_3m_bank',
# 'rate_deu_3m_bank',
],
},
"col_feature_technical_dict": {
"USDJPY": [
'D1-SMA10-val',
'D1-MACD-macd',
'D1-MACD-macdsignal',
'D1-MACD-macdhist',
'D1-ROC2-val',
'D1-MOM4-val',
'D1-RSI10-val',
'D1-BB20-upper',
'D1-BB20-mid',
'D1-BB20-lower',
'D1-CCI20-val',
],
},
},
}
from AlgorithmImports import *
import numpy as np
from datetime import datetime, timedelta
from collections import deque
from indicator_classes import (
CustomPhase,
CustomCRSI,
CustomPivot,
CustomChoppinessIndex,
CustomDX,
CustomMACD,
CustomULTOSC,
)
class SymbolData:
def __init__(
self,
algorithm,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
):
self.symbol = symbol
self.ticker = ticker
self.algorithm = algorithm
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.to_plot = algorithm.to_plot
if general_setting["tickers"][ticker]["type"] in ["forex", "cfd"]:
self.consolidator_type = "quote"
elif general_setting["tickers"][ticker]["type"] in ["equity", "index"]:
self.consolidator_type = "trade"
self.consolidators = {}
for timeframe in self.general_setting["consolidator_timeframes"]:
self.consolidators[timeframe] = DataConsolidator(
algorithm,
symbol,
ticker,
general_setting,
consolidator_settings[timeframe],
indicator_settings,
self.consolidator_type,
)
@property
def IsReady(self):
# All consolidators are ready
is_ready = (
np.prod([self.consolidators[_t].IsReady for _t in self.general_setting["consolidator_timeframes"]]) == 1
)
return is_ready
class DataConsolidator:
def __init__(
self,
algorithm,
symbol,
ticker,
general_setting,
consolidator_setting,
indicator_settings,
consolidator_type,
):
self.symbol = symbol
self.ticker = ticker
self.algorithm = algorithm
self.general_setting = general_setting
self.consolidator_setting = consolidator_setting
self.indicator_settings = indicator_settings
self.consolidator_type = consolidator_type
self.to_plot = algorithm.to_plot
self.indicators = {}
self.ticker_type = self.general_setting["tickers"][self.ticker]["type"]
self.window_multiplier = self.consolidator_setting["window_multiplier_dict"][self.ticker_type]
self.window_length = int(self.consolidator_setting["window"] * self.window_multiplier)
self.time = deque(maxlen=self.window_length)
self.open = deque(maxlen=self.window_length)
self.high = deque(maxlen=self.window_length)
self.low = deque(maxlen=self.window_length)
self.close = deque(maxlen=self.window_length)
self.returns = deque(maxlen=self.window_length)
if "window_large" in self.consolidator_setting:
self.window_length_large = int(self.consolidator_setting["window_large"] * self.window_multiplier)
self.time_large = deque(maxlen=self.window_length_large)
self.close_large = deque(maxlen=self.window_length_large)
self.BarCount = 0
if self.consolidator_type == "quote":
if self.consolidator_setting["timeframe_minutes"] in [5, 15, 30, 60]:
self.Con = QuoteBarConsolidator(
TimeSpan.FromMinutes(self.consolidator_setting["timeframe_minutes"])
)
elif self.consolidator_setting["timeframe_minutes"] in [4 * 60]:
self.Con = QuoteBarConsolidator(self.H4Timer)
elif self.consolidator_setting["timeframe_minutes"] in [24 * 60]:
self.Con = QuoteBarConsolidator(self.D1Timer)
elif self.consolidator_setting["timeframe_minutes"] in [7 * 24 * 60]:
self.Con = QuoteBarConsolidator(self.W1Timer)
elif self.consolidator_type == "trade":
if self.consolidator_setting["timeframe_minutes"] in [5, 15, 30, 60]:
self.Con = TradeBarConsolidator(
TimeSpan.FromMinutes(self.consolidator_setting["timeframe_minutes"])
)
elif self.consolidator_setting["timeframe_minutes"] in [4 * 60]:
self.Con = TradeBarConsolidator(self.H4Timer)
elif self.consolidator_setting["timeframe_minutes"] in [24 * 60]:
self.Con = TradeBarConsolidator(self.D1Timer)
elif self.consolidator_setting["timeframe_minutes"] in [7 * 24 * 60]:
self.Con = TradeBarConsolidator(self.W1Timer)
self.Con.DataConsolidated += self.ConHandler
algorithm.SubscriptionManager.AddConsolidator(symbol, self.Con)
for _indicator in self.consolidator_setting["indicators"]:
self.indicators[_indicator] = self.get_indicator(symbol, _indicator)
def ConHandler(self, sender, bar):
self.BarCount += 1
self.time.appendleft(bar.Time)
self.open.appendleft(bar.Open)
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.close.appendleft(bar.Close)
if "window_large" in self.consolidator_setting:
self.time_large.appendleft(bar.Time)
self.close_large.appendleft(bar.Close)
if len(self.close) > 1:
self.returns.appendleft((self.close[0] / self.close[1]) - 1)
else:
self.returns.appendleft(0)
for _indicator in self.consolidator_setting["indicators"]:
self.update_indicator(bar, _indicator)
def H4Timer(self, dt):
start = (dt if dt.hour > 17 else dt - timedelta(1)).date()
start = datetime.combine(start, datetime.min.time()) + timedelta(hours=17)
return CalendarInfo(start, timedelta(hours=4))
def D1Timer(self, dt):
start = (dt if dt.hour > 17 else dt - timedelta(1)).date()
start = datetime.combine(start, datetime.min.time()) + timedelta(hours=17)
return CalendarInfo(start, timedelta(hours = 12))
def W1Timer(self, dt):
_date = dt.date()
if _date.weekday() == 6:
if dt.hour > 17:
start = _date
else:
start = _date - timedelta(days=_date.weekday()) - timedelta(1)
else:
start = _date - timedelta(days=_date.weekday()) - timedelta(1)
start = datetime.combine(start, datetime.min.time()) + timedelta(hours=17)
return CalendarInfo(start, timedelta(7))
def get_indicator(self, symbol, name):
indicator_setting = self.indicator_settings[name]
indicator_dict = {}
if indicator_setting["type"] == "EMA":
indicator_dict["model"] = ExponentialMovingAverage(
symbol, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "SMA":
indicator_dict["model"] = SimpleMovingAverage(
symbol, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "ROC":
indicator_dict["model"] = RateOfChange(
symbol, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "AO":
indicator_dict["model"] = AwesomeOscillator(
symbol, 10, 20, MovingAverageType.Simple)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "TRIX":
indicator_dict["model"] = Trix(
symbol, indicator_setting["lookback"])
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "MOM":
indicator_dict["model"] = Momentum(
symbol, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "CCI":
indicator_dict["model"] = CommodityChannelIndex(
symbol, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "RSI":
if indicator_setting["ma_type"] == "Exponential":
indicator_dict["model"] = RelativeStrengthIndex(
symbol, indicator_setting["lookback"], MovingAverageType.Exponential
)
elif indicator_setting["ma_type"] == "Simple":
indicator_dict["model"] = RelativeStrengthIndex(
symbol, indicator_setting["lookback"], MovingAverageType.Simple
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "ATR":
indicator_dict["model"] = AverageTrueRange(
symbol, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "PSAR":
indicator_dict["model"] = ParabolicStopAndReverse(symbol, 0.02, 0.02, 0.2)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "BOLL":
if indicator_setting["ma_type"] == "Exponential":
indicator_dict["model"] = BollingerBands(
name,
indicator_setting["lookback"],
indicator_setting["std"],
MovingAverageType.Exponential,
)
elif indicator_setting["ma_type"] == "Simple":
indicator_dict["model"] = BollingerBands(
name,
indicator_setting["lookback"],
indicator_setting["std"],
MovingAverageType.Simple,
)
indicator_dict["upper"] = deque(maxlen=indicator_setting["window"])
indicator_dict["lower"] = deque(maxlen=indicator_setting["window"])
indicator_dict["mid"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "PHASE":
indicator_dict["model"] = CustomPhase(name, indicator_setting["lookback"])
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "CRSI":
indicator_dict["model"] = CustomCRSI(
name,
indicator_setting["rsi_len"],
indicator_setting["rsi_field"],
indicator_setting["rsi_window"],
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "PIVOT":
indicator_dict["model"] = CustomPivot(name, indicator_setting["period"])
elif indicator_setting["type"] == "CHOP":
indicator_dict["model"] = CustomChoppinessIndex(
name, indicator_setting["lookback"]
)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "DX":
indicator_dict["model"] = CustomDX(name, indicator_setting["lookback"])
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "MACD":
indicator_dict["model"] = CustomMACD(name)
indicator_dict["macd"] = deque(maxlen=indicator_setting["window"])
indicator_dict["macdsignal"] = deque(maxlen=indicator_setting["window"])
indicator_dict["macdhist"] = deque(maxlen=indicator_setting["window"])
elif indicator_setting["type"] == "ULTOSC":
indicator_dict["model"] = CustomULTOSC(name)
indicator_dict["val"] = deque(maxlen=indicator_setting["window"])
return indicator_dict
def update_indicator(self, bar, name):
def get_update_val(bar, indicator_setting):
if indicator_setting["field"] == "Close":
val = bar.Close
elif indicator_setting["field"] == "High":
val = bar.High
elif indicator_setting["field"] == "Low":
val = bar.Low
elif indicator_setting["field"] == "Open":
val = bar.Open
return val
indicator_setting = self.indicator_settings[name]
if indicator_setting["type"] in ["EMA","SMA","ROC","MOM","RSI"]:
self.indicators[name]["model"].Update(
bar.EndTime, get_update_val(bar, indicator_setting)
)
self.indicators[name]["val"].appendleft(
self.indicators[name]["model"].Current.Value
)
elif indicator_setting["type"] in ["ATR","PSAR","CCI"]:
self.indicators[name]["model"].Update(bar)
self.indicators[name]["val"].appendleft(
self.indicators[name]["model"].Current.Value
)
elif indicator_setting["type"] in ["AO"]:
# self.indicators[name]["model"].Update(bar)
self.indicators[name]["model"].Update(bar)
self.indicators[name]["val"].appendleft(
self.indicators[name]["model"].Current.Value
)
elif indicator_setting["type"] in ["TRIX"]:
self.indicators[name]["model"].Update(
bar.EndTime, get_update_val(bar, indicator_setting)
)
self.indicators[name]["val"].appendleft(
self.indicators[name]["model"].Current.Value
)
elif indicator_setting["type"] in ["PHASE", "CRSI", "CHOP", "DX", "ULTOSC"]:
self.indicators[name]["model"].Update(bar)
self.indicators[name]["val"].appendleft(
self.indicators[name]["model"].Value
)
elif indicator_setting["type"] in ["PIVOT"]:
self.indicators[name]["model"].Update(bar)
elif indicator_setting["type"] in ["BOLL"]:
self.indicators[name]["model"].Update(
bar.EndTime, get_update_val(bar, indicator_setting)
)
self.indicators[name]["upper"].appendleft(
self.indicators[name]["model"].UpperBand.Current.Value
)
self.indicators[name]["lower"].appendleft(
self.indicators[name]["model"].LowerBand.Current.Value
)
self.indicators[name]["mid"].appendleft(
self.indicators[name]["model"].MiddleBand.Current.Value
)
elif indicator_setting["type"] in ["MACD"]:
self.indicators[name]["model"].Update(bar)
self.indicators[name]["macd"].appendleft(
self.indicators[name]["model"].macd
)
self.indicators[name]["macdsignal"].appendleft(
self.indicators[name]["model"].macdsignal
)
self.indicators[name]["macdhist"].appendleft(
self.indicators[name]["model"].macdhist
)
@property
def IsReady(self):
# All indicators are ready
is_ready = (np.prod([self.indicators[_i]["model"].IsReady for _i in self.indicators]) == 1)
# self.close is fully populated
is_ready = is_ready and (len(self.close) == self.window_length)
if "window_large" in self.consolidator_setting:
# self.close_large is fully populated
is_ready = is_ready and (len(self.close_large) == self.window_length_large)
return is_ready
class MarketHours:
def __init__(self, algorithm, symbol):
self.hours = algorithm.Securities[symbol].Exchange.Hours
self.CurrentOpen = self.hours.GetNextMarketOpen(algorithm.Time, False)
self.CurrentClose = self.hours.GetNextMarketClose(self.CurrentOpen, False)
self.NextOpen = self.hours.GetNextMarketOpen(self.CurrentClose, False)
def Update(self):
self.CurrentOpen = self.NextOpen
self.CurrentClose = self.hours.GetNextMarketClose(self.CurrentOpen, False)
self.NextOpen = self.hours.GetNextMarketOpen(self.CurrentClose, False)
from AlgorithmImports import *
import numpy as np
from datetime import datetime, timedelta
from collections import deque
import talib as ta
import networkx as nx
class CustomSimpleMovingAverage:
def __init__(self, name, lookback):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.queue = deque(maxlen=lookback)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
# Update method is mandatory
def Update(self, EndTime, Val):
self.queue.appendleft(Val)
count = len(self.queue)
self.Time = EndTime
self.Value = sum(self.queue) / count
self.IsReady = count == self.queue.maxlen
def Undo(self):
del self.queue[0]
class CustomAwesomeOscillator:
def __init__(self, name, lookback1, lookback2):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.Value1 = 0
self.Value2 = 0
self.IsReady = False
self.queue1 = deque(maxlen=lookback1)
self.queue2 = deque(maxlen=lookback2)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
# Update method is mandatory
def Update(self, EndTime, Val):
self.queue1.appendleft(Val)
self.queue2.appendleft(Val)
count1 = len(self.queue1)
count2 = len(self.queue2)
self.Time = EndTime
self.Value1 = sum(self.queue1) / count1
self.Value2 = sum(self.queue2) / count2
self.Value = self.Value1 - self.Value2
self.IsReady = ((count1 == self.queue1.maxlen) & (count2 == self.queue2.maxlen) )
def Undo(self):
del self.queue[0]
class CustomPhase:
## Need to check before use!! data is in descending order, not sure whether logic assumes descending/ascending order !!
def __init__(self, name, lookback):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.lookback = lookback
self.close = deque(maxlen=lookback)
self.high = deque(maxlen=lookback)
self.low = deque(maxlen=lookback)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.Time = bar.EndTime
RealPart = 0.0
ImagPart = 0.0
_count = min([len(self.close), len(self.high), len(self.low)])
for J in range(_count):
Weight = (
self.close[J] + self.close[J] + self.high[J] + self.low[J]
) * 10000
if self.lookback != 0:
RealPart = RealPart + np.cos(90 * J / self.lookback) * Weight * 2
ImagPart = (
(ImagPart + np.sin(90 * J / self.lookback) * Weight)
+ (ImagPart + np.sin(180 * J / self.lookback) * Weight)
) / 2
Phase = ((np.arctan(ImagPart / RealPart)) - 0.685) * 100
self.Value = Phase
self.IsReady = _count == self.lookback
class CustomCRSI:
## Need to check before use!! data is in descending order, not sure whether logic assumes descending/ascending order !!
def __init__(self, name, rsi_len, rsi_field, rsi_window):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.rsi_len = rsi_len
self.rsi_field = rsi_field
self.rsi_window = rsi_window
self.RSI = RelativeStrengthIndex(
f"{name}-RSI", rsi_len, MovingAverageType.Exponential
)
self.RSIval = deque(maxlen=rsi_window)
self.CRSIval = deque(maxlen=3)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.Time = bar.EndTime
self.RSI.Update(bar.EndTime, self.get_update_val(bar))
self.RSIval.appendleft(self.RSI.Current.Value)
vibration = 10
torque = 0.618 / (vibration + 1)
phasingLag = (vibration - 1) / 0.618
if len(self.RSIval) > int(phasingLag):
if len(self.CRSIval) > 1:
crsi1 = self.CRSIval[1]
else:
crsi1 = 0
self.Value = (
torque * (2 * self.RSIval[0] - self.RSIval[int(phasingLag)])
+ (1 - torque) * crsi1
)
else:
self.Value = 0
self.CRSIval.appendleft(self.Value)
self.IsReady = (
self.RSI.IsReady
and (len(self.RSIval) == self.rsi_window)
and (len(self.CRSIval) == 3)
)
def get_update_val(self, bar):
if self.rsi_field == "Close":
val = bar.Close
elif self.rsi_field == "High":
val = bar.High
elif self.rsi_field == "Low":
val = bar.Low
elif self.rsi_field == "Open":
val = bar.Open
return val
class CustomPivot:
def __init__(self, name, lookback):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.lookback = lookback
self.close = deque(maxlen=lookback)
self.high = deque(maxlen=lookback)
self.low = deque(maxlen=lookback)
self.arr_time = deque(maxlen=lookback)
self.p = deque(maxlen=lookback)
self.r1 = deque(maxlen=lookback)
self.s1 = deque(maxlen=lookback)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.Time = bar.EndTime
self.arr_time.appendleft(bar.EndTime)
pivotX_Median = (self.high[0] + self.low[0] + self.close[0]) / 3
self.Value = pivotX_Median
self.p.appendleft(pivotX_Median)
self.r1.appendleft(pivotX_Median * 2 - self.low[0])
self.s1.appendleft(pivotX_Median * 2 - self.high[0])
self.IsReady = len(self.p) == self.lookback
class CustomChoppinessIndex:
def __init__(self, name, lookback):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.lookback = lookback
self.high = deque(maxlen=lookback)
self.low = deque(maxlen=lookback)
self.ATR = AverageTrueRange(f"{name}-RSI", 1)
self.ATRval = deque(maxlen=lookback)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.Time = bar.EndTime
self.ATR.Update(bar)
self.ATRval.appendleft(self.ATR.Current.Value)
if (max(self.high) - min(self.low)) == 0:
self.Value = 0
else:
self.Value = (
100
* np.log10(sum(self.ATRval) / (max(self.high) - min(self.low)))
/ np.log10(self.lookback)
)
self.IsReady = self.ATR.IsReady and (len(self.ATRval) == self.lookback)
class CustomDX:
def __init__(self, name, lookback):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.lookback = lookback
self.window_len = lookback * 2 + 25
self.high = deque(maxlen=self.window_len)
self.low = deque(maxlen=self.window_len)
self.close = deque(maxlen=self.window_len)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
ta_out = ta.DX(
np.array(self.high)[::-1],
np.array(self.low)[::-1],
np.array(self.close)[::-1],
timeperiod=self.lookback,
)
self.Value = ta_out[-1]
else:
self.Value = 0
class CustomMACD:
## Although talib assumes data in chronological ascending order, for some reasons, using data in reversed order produces better results
def __init__(self, name):
self.Name = name
self.Time = datetime.min
self.IsReady = False
self.window_len = 100
self.close = deque(maxlen=self.window_len)
self.macd = 0
self.macdsignal = 0
self.macdhist = 0
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
macd, macdsignal, macdhist = ta.MACD(
np.array(self.close), fastperiod=12, slowperiod=26, signalperiod=9
)
self.macd = macd[-1]
self.macdsignal = macdsignal[-1]
self.macdhist = macdhist[-1]
else:
self.macd = 0
self.macdsignal = 0
self.macdhist = 0
#
class CustomMACD:
## Although talib assumes data in chronological ascending order, for some reasons, using data in reversed order produces better results
def __init__(self, name):
self.Name = name
self.Time = datetime.min
self.IsReady = False
self.window_len = 100
self.close = deque(maxlen=self.window_len)
self.macd = 0
self.macdsignal = 0
self.macdhist = 0
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
macd, macdsignal, macdhist = ta.MACD(
np.array(self.close), fastperiod=12, slowperiod=26, signalperiod=9
)
self.macd = macd[-1]
self.macdsignal = macdsignal[-1]
self.macdhist = macdhist[-1]
else:
self.macd = 0
self.macdsignal = 0
self.macdhist = 0
class CustomULTOSC:
def __init__(self, name):
self.Name = name
self.Time = datetime.min
self.Value = 0
self.IsReady = False
self.window_len = 100
self.high = deque(maxlen=self.window_len)
self.low = deque(maxlen=self.window_len)
self.close = deque(maxlen=self.window_len)
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}. Value: {3}".format(
self.Name, self.IsReady, self.Time, self.Value
)
def Update(self, bar):
self.high.appendleft(bar.High)
self.low.appendleft(bar.Low)
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
ta_out = ta.ULTOSC(
np.array(self.high)[::-1],
np.array(self.low)[::-1],
np.array(self.close)[::-1],
timeperiod1=7,
timeperiod2=14,
timeperiod3=28,
)
self.Value = ta_out[-1]
else:
self.Value = 0
class CustomVMD:
def __init__(self, name):
self.Name = name
self.Time = datetime.min
self.IsReady = False
self.window_len = 120
self.alpha = 2000
self.k = 4
self.close = deque(maxlen=self.window_len)
self.Values = {}
for j in range(self.k):
self.Values[f"c{j}"] = 0
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}".format(
self.Name, self.IsReady, self.Time
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.window_len)
if self.IsReady:
tsv = np.array(self.close)[::-1]
tso = VMD(tsv, self.k, alpha=self.alpha, tau=0, DC=False, init=1, tol=1e-7)
for j in range(self.k):
self.Values[f"c{j}"] = tso[-1,j]
class CustomVG:
def __init__(self, name, lookback, centrality_type):
self.Name = name
self.Time = datetime.min
self.IsReady = False
self.lookback = lookback
self.centrality_type = centrality_type
self.max_k = 4
self.centrality_type_map_dict = {
'degree': nx.degree_centrality,
'closeness': nx.closeness_centrality,
'harmonic': nx.harmonic_centrality,
}
self.close = deque(maxlen=lookback)
self.Values = {}
for j in range(self.max_k):
self.Values[f"c{j}"] = 0
def __repr__(self):
return "{0} -> IsReady: {1}. Time: {2}".format(
self.Name, self.IsReady, self.Time
)
def Update(self, bar):
self.close.appendleft(bar.Close)
self.Time = bar.EndTime
self.IsReady = len(self.close) == (self.lookback)
if self.IsReady:
tsv = np.array(self.close)[::-1]
tsg = visibility_graph(tsv)
for j in range(self.max_k):
tsg_shell = nx.k_shell(tsg, j+1)
centralities = list(self.centrality_type_map_dict[self.centrality_type](tsg_shell).values())
self.Values[f"c{j}"] = np.mean(centralities)
# Version
# Forex LSTM V2.0 Oanada
##-##
IS_LIVE = False
IS_PAPER = True
# notification_value = 'PAPER'
TO_SAVE_DATA = False
# if IS_LIVE:
# TO_SAVE_DATA = False
# else:
# TO_SAVE_DATA = True
##-##
from AlgorithmImports import *
import numpy as np
import pandas as pd
from datetime import datetime, timedelta
from collections import deque
import pickle
import torch
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(f"device: {device}")
from config import (
general_setting,
consolidator_settings,
indicator_settings,
signal_settings,
model_settings,
)
from data_classes import (
SymbolData,
MarketHours,
)
from signal_classes import (
FxLstmSignal,
)
from model_functions import (
get_threshold,
set_seed,
)
from model_classes_both import get_torch_rnn_dataloaders as get_torch_rnn_dataloaders_both
from model_classes_both import get_rnn_model as get_rnn_model_both
from model_classes_both import get_predictions as get_predictions_both
from model_classes_hybrid import get_torch_rnn_dataloaders as get_torch_rnn_dataloaders_hybrid
from model_classes_hybrid import get_rnn_model as get_rnn_model_hybrid
from model_classes_hybrid import get_predictions as get_predictions_hybrid
from model_classes_hybrid import get_regression_pred_decision, get_prediction_hybrid_regression
from sklearn.metrics import mean_squared_error
signal_mapping = {
"FxLstm_Hybrid_USDJPY_Trail": FxLstmSignal,
}
from QuantConnect.DataSource import *
class FxLstmAlgo(QCAlgorithm):
def Initialize(self):
# self.SetTimeZone(TimeZones.Johannesburg)
self.set_time_zone(TimeZones.NEW_YORK)
IS_LIVE = self.get_parameter("IS_LIVE")
IS_PAPER = self.get_parameter("IS_PAPER")
if IS_LIVE == 'True':
IS_LIVE = True
else:
IS_LIVE = False
if IS_PAPER == 'True':
IS_PAPER = True
else:
IS_PAPER = False
if IS_PAPER:
self.notification_value = 'PAPER'
else:
self.notification_value = 'REAL'
if TO_SAVE_DATA:
self.SetStartDate(2013, 1, 1)
else:
self.SetStartDate(2024, 1, 1) # SR: 1.332
# self.set_end_date(2024, 1 ,10)
self.SetCash(370000)
if TO_SAVE_DATA:
self.SetWarmUp(int(12 * 20 * 24 * 60), Resolution.Minute)
else:
self.SetWarmUp(int(6 * 20 * 24 * 60), Resolution.Minute)
self.SetBrokerageModel(BrokerageName.OandaBrokerage, AccountType.Margin)
self.to_plot = False
self.general_setting = general_setting
self.consolidator_settings = consolidator_settings
self.indicator_settings = indicator_settings
self.signal_settings = signal_settings
self.model_settings = model_settings
self.ref_ticker = "USDJPY"
self.model_name = general_setting["model_name"]
self.month_start_date = None
self.IS_LIVE = IS_LIVE
self.TO_SAVE_DATA = TO_SAVE_DATA
self.signals = self.general_setting["signals"]
self.lstm_tickers = self.general_setting["lstm_tickers"]
self.last_loop_minute = None
self.counter = 0
self.mdd = 0
self.CanTrade = True
self.peak_value = 0
self.his_port_value = {}
self.roll_mdd = 0
self.add_equity('UUP', resolution = Resolution.DAILY)
self.UUP__macd = self.macd("UUP", 12, 26, 9, MovingAverageType.EXPONENTIAL, Resolution.DAILY)
self.UUP__trix = self.trix("UUP", 20)
self.UUP__roc = self.roc("UUP", 10)
self.UUP = {}
self.prediction_dict = {}
for _signal in self.signals:
self.prediction_dict[_signal] = {}
for ticker in self.lstm_tickers:
if ticker in self.signal_settings[_signal]['valid_tickers']:
self.prediction_dict[_signal][ticker] = 1
# Data Initialization
self.Data = {}
self.Signal = {}
self.Counter = {}
self.SymbolMarketHours = {}
self.symbol_ticker_map = {}
self.ticker_symbol_map = {}
self.output_data_dict = {}
self.Futures = {}
self.FuturesTracker = {}
self.FuturesSymbol = {}
self.FuturesRefSymbol = {}
self.entry_ticket = None
self.tp1_ticket = []
self.tp2_ticket = []
self.sl_ticket = []
self.Margin_Call = False
self.overweekend = 0
self.leftover_qty = None
self.leftover_ticker = None
self.prev_equity = None
self.freeze = 0
self.SPY_equity = self.add_equity("SPY", Resolution.MINUTE)
for ticker in self.general_setting["tickers"]:
if general_setting["tickers"][ticker]["type"] == "equity":
symbol = self.AddEquity(
ticker,
Resolution.Minute,
dataNormalizationMode=DataNormalizationMode.Raw,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "forex":
symbol = self.AddForex(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
elif general_setting["tickers"][ticker]["type"] == "cfd":
symbol = self.AddCfd(
ticker,
Resolution.Minute,
Market.Oanda,
).Symbol
self.Data[symbol] = SymbolData(
self,
symbol,
ticker,
general_setting,
consolidator_settings,
indicator_settings,
)
self.Counter[symbol] = {}
self.Counter[symbol]["counter"] = 0
self.Counter[symbol]["last_order_counter"] = 0
self.SymbolMarketHours[symbol] = MarketHours(self, symbol)
self.symbol_ticker_map[symbol] = ticker
self.ticker_symbol_map[ticker] = symbol
self.Signal[symbol] = {}
for _signal in self.signals:
self.Signal[symbol][_signal] = signal_mapping[_signal](
self, symbol, ticker, self.general_setting, self.signal_settings[_signal]
)
# Model Initialization
self.Models = {}
self.Scalers = {}
self.ModelParams = {}
for lstm_ticker in self.lstm_tickers:
self.Models[lstm_ticker] = {}
self.Scalers[lstm_ticker] = {}
self.ModelParams[lstm_ticker] = {}
for model_type in self.general_setting["model_types"]:
self.Models[lstm_ticker][model_type] = {}
self.Scalers[lstm_ticker][model_type] = {}
self.ModelParams[lstm_ticker][model_type] = {}
if model_type == 'both':
self.Models[lstm_ticker][model_type]['both'] = None
self.Scalers[lstm_ticker][model_type]['both'] = None
if model_type == 'hybrid':
self.Models[lstm_ticker][model_type]['fundamental'] = None
self.Models[lstm_ticker][model_type]['technical'] = None
self.Scalers[lstm_ticker][model_type]['fundamental'] = None
self.Scalers[lstm_ticker][model_type]['technical'] = None
# Model Data Initialization
self.data_list_tickers = {}
self.has_initialized_model_data = {}
if self.TO_SAVE_DATA:
self.ModelData = {}
else:
self.ModelData = pickle.loads(bytes(self.ObjectStore.ReadBytes(f"MODEL_DATA_{self.model_name}")))
for lstm_ticker in self.lstm_tickers:
self.data_list_tickers[lstm_ticker] = []
self.has_initialized_model_data[lstm_ticker] = False
if lstm_ticker not in self.ModelData:
self.ModelData[lstm_ticker] = pd.DataFrame()
self.Schedule.On(
self.DateRules.EveryDay(self.SPY_equity.symbol),
self.TimeRules.AfterMarketOpen(self.SPY_equity.symbol),
self.Start_Of_Day,
)
self.Schedule.On(
self.DateRules.EveryDay(self.SPY_equity.symbol),
self.TimeRules.At(self.general_setting['FxLstm_prediction_hour'], 5, 0),
self.Prepare_Model_Data,
)
self.Schedule.On(
self.DateRules.MonthStart(self.SPY_equity.symbol),
self.TimeRules.AfterMarketOpen(self.SPY_equity.symbol, 1),
self.Get_Month_Start_Date,
)
self.Schedule.On(
self.DateRules.EveryDay(self.SPY_equity.symbol),
self.TimeRules.AfterMarketOpen(self.SPY_equity.symbol, 2),
self.Train_Model_Both,
)
self.Schedule.On(
self.DateRules.EveryDay(self.SPY_equity.symbol),
self.TimeRules.AfterMarketOpen(self.SPY_equity.symbol, 3),
self.Train_Model_Hybrid,
)
self.external_data = {}
for _dn in self.general_setting["external_data"]:
self.external_data[_dn] = {}
self.external_data[_dn]['time'] = None
self.external_data[_dn]['value'] = None
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
self.Log(f"{str(self.Time)}: {_dn}: Loading Initial GSheet Data")
link = self.general_setting["external_data"][_dn]['link']
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
to_run = True
while to_run:
try:
data = self.Download(link)
rows = []
for row in data.split('\n'):
content = row.replace('\r', '').lower().split(',')
if len(content) == 2:
rows.append(content)
data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
data_df[col_val] = data_df[col_val].astype(float)
self.external_data[_dn]['data'] = data_df.copy()
to_run = False
except:
pass
self.Log(f"{str(self.Time)}: {_dn}: Initial GSheet Data Loaded")
if source == 'NasdaqDataLink':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(NasdaqDataLink, ref, Resolution.Daily).Symbol
if source == 'equity':
ticker = self.general_setting["external_data"][_dn]['ticker']
self.external_data[_dn]['symbol'] = self.AddEquity(ticker, Resolution.Daily).Symbol
if source == 'USTreasuryYieldCurveRate':
ref = self.general_setting["external_data"][_dn]['ref']
self.external_data[_dn]['symbol'] = self.AddData(USTreasuryYieldCurveRate, ref).Symbol
def Start_Of_Day(self):
# if self.IS_LIVE and (not self.IsWarmingUp):
# for _dn in self.general_setting["external_data"]:
# source = self.general_setting["external_data"][_dn]['source']
# if source == 'gsheet':
# self.Log(f"{str(self.Time)}: {_dn}: Loading GSheet Data")
# link = self.general_setting["external_data"][_dn]['link']
# col_date = self.general_setting["external_data"][_dn]['col_date']
# col_val = self.general_setting["external_data"][_dn]['col_val']
# to_run = True
# while to_run:
# try:
# data = self.Download(link)
# rows = []
# for row in data.split('\n'):
# content = row.replace('\r', '').lower().split(',')
# if len(content) == 2:
# rows.append(content)
# data_df = pd.DataFrame(np.array(rows[1:]), columns=rows[0])
# data_df[col_date] = data_df[col_date].apply(lambda s: datetime.strptime(s, '%Y-%m-%d'))
# data_df[col_val] = data_df[col_val].astype(float)
# self.external_data[_dn]['data'] = data_df.copy()
# to_run = False
# except:
# pass
# self.Log(f"{str(self.Time)}: {_dn}: GSheet Data Loaded")
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'gsheet':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
lag_days = self.general_setting["external_data"][_dn]['lag_days']
data = self.external_data[_dn]['data'][self.external_data[_dn]['data'][col_date] < (self.Time - timedelta(days=lag_days))]
if len(data) > 0:
self.external_data[_dn]['time'] = data[col_date].values[-1]
self.external_data[_dn]['value'] = data[col_val].values[-1]
if source == 'USTreasuryYieldCurveRate':
col_date = self.general_setting["external_data"][_dn]['col_date']
col_val = self.general_setting["external_data"][_dn]['col_val']
symbol = self.external_data[_dn]['symbol']
history = self.History(USTreasuryYieldCurveRate, symbol, 1, Resolution.Daily)
history = history.reset_index()
if len(history) > 0:
if col_val in history.columns:
self.external_data[_dn]['time'] = pd.to_datetime(history[col_date], utc=True).iloc[0].replace(tzinfo=None)
self.external_data[_dn]['value'] = history[col_val].values[0]
def Prepare_Model_Data(self):
# self.Log(f"{str(self.Time)}: Preparing Model Data")
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
cols_data = self.model_settings['cols_data']
col_fundamental = self.model_settings['col_fundamental']
col_technical = self.model_settings['col_technical']
start_year = self.model_settings['start_year']
trade_hour = self.model_settings['trade_hour']
scaled_tickers = self.model_settings['scaled_tickers']
prediction_lookforward_days = self.model_settings['prediction_lookforward_days']
col_target_gains = f"gains_N{self.model_settings['prediction_lookforward_days']}D"
inflation_map_dict = self.model_settings['inflation_map_dict']
to_save_data = False
for lstm_ticker in self.lstm_tickers:
data_df = self.ModelData[lstm_ticker].copy()
has_new_data = False
if len(self.data_list_tickers[lstm_ticker]) > 0:
has_new_data = True
data_df_new = pd.DataFrame(self.data_list_tickers[lstm_ticker]).copy()
if lstm_ticker in scaled_tickers:
data_df_new[col_price] = data_df_new[col_price] / 100
data_df_new[col_price_cur] = data_df_new[col_price_cur] / 100
data_df_new = data_df_new[cols_data]
data_df_new['year'] = data_df_new['datetime'].dt.year
data_df_new['hour'] = data_df_new['datetime'].dt.hour
data_df_new['month'] = data_df_new['datetime'].dt.month
data_df_new['year_month'] = data_df_new['year'].astype(str) + "-" + data_df_new['month'].astype(str).apply(lambda s: s.zfill(2))
data_df = pd.concat([data_df, data_df_new])
if len(data_df) > 0:
if (not self.has_initialized_model_data[lstm_ticker]) or has_new_data:
self.has_initialized_model_data[lstm_ticker] = True
data_df.reset_index(drop=True, inplace=True)
if self.TO_SAVE_DATA:
data_df.drop_duplicates('datetime', keep='last', inplace=True)
else:
data_df.drop_duplicates('datetime', keep='first', inplace=True)
data_df.reset_index(drop=True, inplace=True)
data_df.sort_values('datetime', ascending=True, inplace=True)
data_df.reset_index(drop=True, inplace=True)
for col in col_fundamental + col_technical:
data_df[col] = data_df[col].fillna(method='ffill')
data_df = data_df[data_df['year'] >= start_year]
data_df = data_df[data_df['hour'] == trade_hour]
data_df.reset_index(drop=True, inplace=True)
for col in inflation_map_dict:
col_cpi = inflation_map_dict[col]
## FRED CPI value is contribution to inflation. To test using it directly without differencing, as well as d1, d30 difference on BOTH
data_df[f"{col}_d1"] = data_df[col_cpi] - data_df[col_cpi].shift(1)
data_df[f"{col}_d30"] = data_df[col_cpi] - data_df[col_cpi].shift(1)
data_df[col_target_gains] = data_df[col_price].shift(-prediction_lookforward_days) - data_df[col_price]
self.ModelData[lstm_ticker] = data_df.copy()
self.data_list_tickers[lstm_ticker] = []
to_save_data = True
if to_save_data and self.TO_SAVE_DATA:
self.ObjectStore.SaveBytes(f"MODEL_DATA_{self.model_name}", pickle.dumps(self.ModelData))
self.Log(f"{str(self.Time)}: Model Data Saved At: MODEL_DATA_{self.model_name}")
# self.Log(f"{str(self.Time)}: Model Data Prepared")
def Get_Month_Start_Date(self):
self.month_start_date = self.Time
def Train_Model_Both(self):
model_type = "both"
# self.Log(f"{str(self.Time)}: {model_type}: Training Model")
model_setting = self.model_settings[f"model_settings_{model_type}"]
col_date = self.model_settings['col_date']
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
col_target = self.model_settings['col_target']
prediction_lookforward_days = self.model_settings['prediction_lookforward_days']
col_target_gains = f"gains_N{self.model_settings['prediction_lookforward_days']}D"
use_gru_model = model_setting['use_gru_model']
use_dual_lstm = model_setting['use_dual_lstm']
epochs = model_setting['epochs']
hidden_size = model_setting['hidden_size']
window_size = model_setting['window_size']
thres_multiplier = model_setting['thres_multiplier']
use_early_stop = model_setting['use_early_stop']
learning_rate = model_setting['learning_rate']
batch_size = model_setting['batch_size']
use_weighted_sampler = model_setting['use_weighted_sampler']
volatility_type = model_setting['volatility_type']
valid_lookback_months = model_setting['valid_lookback_months']
train_lookback_months = model_setting['train_lookback_months']
inflation_map_dict = self.model_settings['inflation_map_dict']
for lstm_ticker in self.lstm_tickers:
if self.month_start_date is None:
continue
else:
month_start_year = self.month_start_date.year
month_start_month = self.month_start_date.month
month_start_day = self.month_start_date.day
model_train_day = month_start_day + self.general_setting["lstm_model_training_displace_days"][lstm_ticker]
to_train = (self.Time.year == month_start_year) and (self.Time.month == month_start_month) and (self.Time.day == model_train_day)
if not to_train:
continue
data_df = self.ModelData[lstm_ticker].copy()
if len(data_df) == 0:
continue
col_feature_both = model_setting["col_feature_dict"][lstm_ticker]
year_month_list = sorted(list(set(data_df['year_month'])))
year_month_vec = np.array(year_month_list)
year_vec = np.array(sorted(list(set(data_df['year']))))
test_year_month = f"{self.Time.year}-{str(self.Time.month).zfill(2)}"
valid_year_month_list = list(year_month_vec[year_month_vec < test_year_month][-valid_lookback_months:])
if len(valid_year_month_list) < valid_lookback_months:
continue
if np.sum(year_month_vec < min(valid_year_month_list)) == 0:
continue
train_year_month_list = list(year_month_vec[year_month_vec < min(valid_year_month_list)][-train_lookback_months:])
if len(train_year_month_list) < train_lookback_months:
continue
data_df_temp = data_df.copy()
if volatility_type == 'thres_v1':
col_target_gains_thres = 0.00200
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
if volatility_type == 'thres_v2':
col_target_gains_thres = 0.00235
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
if volatility_type == 'thres_auto_v1':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
if volatility_type == 'thres_auto_v2':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list+valid_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
data_df_temp[col_target] = 1
data_df_temp.loc[data_df_temp[col_target_gains] < -col_target_gains_thres, col_target] = 0
data_df_temp.loc[data_df_temp[col_target_gains] > col_target_gains_thres, col_target] = 2
self.ModelParams[lstm_ticker][model_type]['col_target_gains_thres'] = col_target_gains_thres
data_df_temp = data_df_temp.dropna()
data_df_temp.reset_index(drop=True, inplace=True)
train_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
valid_df = data_df_temp[data_df_temp['year_month'].isin(valid_year_month_list)].copy()
valid_df_windowed = pd.concat([train_df,valid_df]).copy()
valid_df_windowed = valid_df_windowed.tail(len(valid_df) + window_size-1)
set_seed(100)
(train_loader, val_loader,
_, scaler, weighted_sampler, class_weights) = get_torch_rnn_dataloaders_both(
[col_price] + col_feature_both, col_target, train_df, valid_df_windowed, None, window_size, batch_size,
use_weighted_sampler=use_weighted_sampler,
has_test_data=False,
)
self.Scalers[lstm_ticker][model_type]['both'] = None
self.Scalers[lstm_ticker][model_type]['both'] = scaler
self.Models[lstm_ticker][model_type]['both'] = None
self.Models[lstm_ticker][model_type]['both'] = get_rnn_model_both(
[col_price] + col_feature_both, train_loader, val_loader,
epochs, batch_size, learning_rate, window_size, hidden_size, device,
use_early_stop=use_early_stop, use_weighted_sampler=use_weighted_sampler, class_weights=class_weights,
use_dual_lstm=use_dual_lstm, use_gru_model=use_gru_model,
)
self.Log(f"{str(self.Time)}: {model_type}: {lstm_ticker}: Model Trained")
def Train_Model_Hybrid(self):
model_type = "hybrid"
# self.Log(f"{str(self.Time)}: {model_type}: Training Model")
model_setting = self.model_settings[f"model_settings_{model_type}"]
col_date = self.model_settings['col_date']
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
col_target = self.model_settings['col_target']
prediction_lookforward_days = self.model_settings['prediction_lookforward_days']
col_target_gains = f"gains_N{self.model_settings['prediction_lookforward_days']}D"
use_gru_model = model_setting['use_gru_model']
use_dual_lstm = model_setting['use_dual_lstm']
epochs = model_setting['epochs']
hidden_size = model_setting['hidden_size']
window_size = model_setting['window_size']
thres_multiplier = model_setting['thres_multiplier']
learning_rate = model_setting['learning_rate']
batch_size = model_setting['batch_size']
volatility_type = model_setting['volatility_type']
valid_lookback_months = model_setting['valid_lookback_months']
train_lookback_months = model_setting['train_lookback_months']
inflation_map_dict = self.model_settings['inflation_map_dict']
for lstm_ticker in self.lstm_tickers:
if self.month_start_date is None:
# self.debug('condition 0 failed')
continue
else:
month_start_year = self.month_start_date.year
month_start_month = self.month_start_date.month
month_start_day = self.month_start_date.day
model_train_day = month_start_day + self.general_setting["lstm_model_training_displace_days"][lstm_ticker]
to_train = (self.Time.year == month_start_year) and (self.Time.month == month_start_month) and (self.Time.day == model_train_day)
if not to_train:
continue
data_df = self.ModelData[lstm_ticker]
if len(data_df) == 0:
continue
col_feature_fundamental = model_setting["col_feature_fundamental_dict"][lstm_ticker]
col_feature_technical = model_setting["col_feature_technical_dict"][lstm_ticker]
year_month_list = sorted(list(set(data_df['year_month'])))
year_month_vec = np.array(year_month_list)
year_vec = np.array(sorted(list(set(data_df['year']))))
test_year_month = f"{self.Time.year}-{str(self.Time.month).zfill(2)}"
valid_year_month_list = list(year_month_vec[year_month_vec < test_year_month][-valid_lookback_months:])
if len(valid_year_month_list) < valid_lookback_months:
continue
if np.sum(year_month_vec < min(valid_year_month_list)) == 0:
continue
train_year_month_list = list(year_month_vec[year_month_vec < min(valid_year_month_list)][-train_lookback_months:])
if len(train_year_month_list) < train_lookback_months:
continue
data_df_temp = data_df.copy()
if volatility_type == 'thres_v1':
col_target_gains_thres = 0.00200
if volatility_type == 'thres_v2':
col_target_gains_thres = 0.00235
if volatility_type == 'thres_auto_v1':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
if volatility_type == 'thres_auto_v2':
thres_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list+valid_year_month_list)].copy()
thres_df.reset_index(drop=True, inplace=True)
col_target_gains_thres = get_threshold(thres_df[col_price]) * thres_multiplier
self.ModelParams[lstm_ticker][model_type]['col_target_gains_thres'] = col_target_gains_thres
data_df_temp[col_target] = data_df_temp[col_price].shift(-prediction_lookforward_days)
data_df_temp = data_df_temp.dropna()
data_df_temp.reset_index(drop=True, inplace=True)
train_df = data_df_temp[data_df_temp['year_month'].isin(train_year_month_list)].copy()
valid_df = data_df_temp[data_df_temp['year_month'].isin(valid_year_month_list)].copy()
# self.Log(f"last date of train is {data_df_temp}")
set_seed(100)
(train_loader, val_loader, _, scaler) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_fundamental, col_target, train_df, valid_df, None, window_size, batch_size,
has_test_data=False,
)
self.Scalers[lstm_ticker][model_type]['fundamental'] = None
self.Scalers[lstm_ticker][model_type]['fundamental'] = scaler
self.Models[lstm_ticker][model_type]['fundamental'] = None
self.Models[lstm_ticker][model_type]['fundamental'] = get_rnn_model_hybrid(
[col_price] + col_feature_fundamental, train_loader, val_loader,
epochs, learning_rate, hidden_size, device,
use_dual_lstm=use_dual_lstm, use_gru_model=use_gru_model,
)
y_pred_val = get_predictions_hybrid(
val_loader,
self.Models[lstm_ticker][model_type]['fundamental'],
self.Scalers[lstm_ticker][model_type]['fundamental'],
[col_price] + col_feature_fundamental,
device,
)
valid_df['pred_price_fundamental'] = y_pred_val
valid_df['pred_fundamental'] = (valid_df['pred_price_fundamental'] - valid_df[col_price]).apply(get_regression_pred_decision, col_target_gains_thres=col_target_gains_thres)
set_seed(100)
(train_loader, val_loader, _, scaler) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_technical, col_target, train_df, valid_df, None, window_size, batch_size,
has_test_data=False,
)
self.Scalers[lstm_ticker][model_type]['technical'] = scaler
self.Models[lstm_ticker][model_type]['technical'] = get_rnn_model_hybrid(
[col_price] + col_feature_technical, train_loader, val_loader,
epochs, learning_rate, hidden_size, device,
use_dual_lstm=use_dual_lstm, use_gru_model=use_gru_model,
)
y_pred_val = get_predictions_hybrid(
val_loader,
self.Models[lstm_ticker][model_type]['technical'],
self.Scalers[lstm_ticker][model_type]['technical'],
[col_price] + col_feature_technical,
device,
)
valid_df['pred_price_technical'] = y_pred_val
valid_df['pred_technical'] = (valid_df['pred_price_technical'] - valid_df[col_price]).apply(get_regression_pred_decision, col_target_gains_thres=col_target_gains_thres)
fundamental_mse = mean_squared_error(valid_df['pred_price_fundamental'], valid_df[col_target])
technical_mse = mean_squared_error(valid_df['pred_price_technical'], valid_df[col_target])
self.ModelParams[lstm_ticker][model_type]['fundamental_mse'] = fundamental_mse
self.ModelParams[lstm_ticker][model_type]['technical_mse'] = technical_mse
self.Log(f"{str(self.Time)}: {model_type}: {lstm_ticker}: Model Trained")
def OnData(self, data):
# Prepare external Data
for _dn in self.general_setting["external_data"]:
source = self.general_setting["external_data"][_dn]['source']
if source == 'NasdaqDataLink':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
self.external_data[_dn]['value'] = data[symbol].Value
if source == 'equity':
symbol = self.external_data[_dn]['symbol']
if data.ContainsKey(symbol):
if data[symbol] is not None:
self.external_data[_dn]['time'] = data[symbol].Time
self.external_data[_dn]['value'] = data[symbol].Price
# if self.IS_LIVE:
# if self.mdd >= 0.2:
# self.CanTrade = False
# self.counter = 0
# self.liquidate()
# self.peak_value = 0
# self.mdd = 0
# self.Log("USDJPY Stop Trading")
# a = { "text": f"[USDJPY Real Money DONOT TRADE update] USDJPY Stop Trading" }
# payload = json.dumps(a)
# self.notify.web("https://hooks.slack.com/services/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", payload)
# if self.counter != 216:
# self.counter += 1
# self.peak_value = max(self.peak_value, self.portfolio.total_portfolio_value)
# self.mdd = max(self.mdd, 1- float(self.portfolio.total_portfolio_value / self.peak_value) )
# else:
# self.counter = 0
# self.peak_value = 0
# self.mdd = 0
if self.CanTrade == False:
return
if self.IS_LIVE and self.IsWarmingUp == False :
self.his_port_value[self.Time] = self.portfolio.total_portfolio_value
curr_ports = [self.his_port_value[x] for x in self.his_port_value.keys() if x>=self.time-timedelta(hours=36)]
curr_peak = max(curr_ports)
self.roll_mdd = (1-self.portfolio.total_portfolio_value/curr_peak)
if self.roll_mdd>= 0.12:
self.CanTrade = False
self.Log("USDJPY Stop Trading")
a = { "text": f"[[US--{self.notification_value}--USDJPY DONOT TRADE update] USDJPY Stop Trading" }
payload = json.dumps(a)
self.notify.web("https://hooks.slack.com/services/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", payload)
if self.CanTrade:
if (self.Time.minute == 0) or (self.Time.minute == 10) or (self.Time.minute == 20) or (self.Time.minute == 30) or (self.Time.minute == 40) or (self.Time.minute == 50):
if (not self.portfolio['USDJPY'].invested) and self.prev_equity:
if self.portfolio.total_portfolio_value < self.prev_equity:
self.freeze = 1
self.Log("Start to freeze")
a = { "text": f"[US--{self.notification_value}--USDJPY freeze update] Start to freeze" }
payload = json.dumps(a)
self.notify.web("https://hooks.slack.com/services/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", payload)
self.prev_equity = None
if self.freeze != 0 :
if self.Time.hour == 18:
# if self.freeze == 13 :
self.freeze = 0
self.Log("Freeze is over")
a = { "text": f"[US--{self.notification_value}--USDJPY freeze update] USDJPY freeze period over" }
payload = json.dumps(a)
self.notify.web("https://hooks.slack.com/services/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", payload)
else:
# self.freeze +=1
return
else:
for symbol, symbolData in self.Data.items():
if not (
data.ContainsKey(symbol)
and data[symbol] is not None
and symbolData.IsReady
):
continue
ticker = self.symbol_ticker_map[symbol]
is_valid_time = self.Time.minute == 0
is_valid_time = is_valid_time and (self.Time.hour in [self.general_setting['FxLstm_prediction_hour']])
if is_valid_time:
if ticker in self.lstm_tickers:
data_dict = {}
data_dict["datetime"] = self.Time
data_dict["price"] = np.round(data[symbol].Price, 6)
# Daily Data
_consolidator = symbolData.consolidators["D1"]
data_dict["close_D1"] = _consolidator.close[0]
# External Data
for _dn in self.general_setting["external_data"]:
data_dict[_dn] = self.external_data[_dn]['value']
# Technical Features
for _tf in self.general_setting["features"]:
_consolidator = symbolData.consolidators[_tf]
for _in in self.general_setting["features"][_tf]:
_indicator = _consolidator.indicators[_in]
if _in in self.general_setting["features_val_map"]:
for _v in self.general_setting["features_val_map"][
_in
]:
data_dict[f"{_tf}-{_in}-{_v}"] = np.round(
_indicator[_v][0], 5
)
if self.TO_SAVE_DATA:
if not self.IsWarmingUp:
self.data_list_tickers[ticker] += [data_dict]
else:
self.data_list_tickers[ticker] += [data_dict]
col_price = self.model_settings['col_price']
col_price_cur = self.model_settings['col_price_cur']
cols_data = self.model_settings['cols_data']
col_fundamental = self.model_settings['col_fundamental']
col_technical = self.model_settings['col_technical']
start_year = self.model_settings['start_year']
trade_hour = self.model_settings['trade_hour']
col_target = self.model_settings['col_target']
scaled_tickers = self.model_settings['scaled_tickers']
inflation_map_dict = self.model_settings['inflation_map_dict']
max_window_size = self.model_settings['max_window_size']
test_df = pd.DataFrame()
if len(self.data_list_tickers[ticker]) > 0:
data_df_new = pd.DataFrame(self.data_list_tickers[ticker]).copy()
if ticker in scaled_tickers:
data_df_new[col_price] = data_df_new[col_price] / 100
data_df_new[col_price_cur] = data_df_new[col_price_cur] / 100
data_df_new = data_df_new[cols_data]
data_df_new['year'] = data_df_new['datetime'].dt.year
data_df_new['hour'] = data_df_new['datetime'].dt.hour
data_df_new['month'] = data_df_new['datetime'].dt.month
data_df_new['year_month'] = data_df_new['year'].astype(str) + "-" + data_df_new['month'].astype(str).apply(lambda s: s.zfill(2))
data_df = self.ModelData[ticker].copy()
if len(data_df) > 0:
idx = data_df['datetime'] < self.Time
data_df = data_df[idx]
data_df.reset_index(drop=True, inplace=True)
data_df = pd.concat([data_df, data_df_new])
data_df.reset_index(drop=True, inplace=True)
data_df.drop_duplicates('datetime', keep='last', inplace=True)
data_df.reset_index(drop=True, inplace=True)
data_df.sort_values('datetime', ascending=True, inplace=True)
data_df.reset_index(drop=True, inplace=True)
for col in col_fundamental + col_technical:
data_df[col] = data_df[col].fillna(method='ffill')
data_df = data_df[data_df['year'] >= start_year]
data_df = data_df[data_df['hour'] == trade_hour]
data_df.reset_index(drop=True, inplace=True)
for col in inflation_map_dict:
col_cpi = inflation_map_dict[col]
## FRED CPI value is contribution to inflation. To test using it directly without differencing, as well as d1, d30 difference on BOTH
data_df[f"{col}_d1"] = data_df[col_cpi] - data_df[col_cpi].shift(1)
data_df[f"{col}_d30"] = data_df[col_cpi] - data_df[col_cpi].shift(1)
test_df = data_df.tail(max_window_size).copy()
test_df.reset_index(drop=True, inplace=True)
for model_type in self.general_setting["model_types"]:
if len(test_df) == 0:
continue
if model_type == 'both':
if self.Models[ticker][model_type]['both'] is None:
continue
if model_type == 'hybrid':
if self.Models[ticker][model_type]['fundamental'] is None:
continue
if self.Models[ticker][model_type]['technical'] is None:
continue
model_setting = self.model_settings[f"model_settings_{model_type}"]
test_df_windowed = test_df.tail(model_setting['window_size']).copy()
test_df_windowed.reset_index(drop=True, inplace=True)
if len(test_df_windowed) != model_setting['window_size']:
continue
if model_type == 'both':
col_feature_both = model_setting["col_feature_dict"][ticker]
test_df_windowed[col_target] = 1
(_, _, test_loader, _, _, _) = get_torch_rnn_dataloaders_both(
[col_price] + col_feature_both, col_target, None, None, test_df_windowed.copy(),
model_setting['window_size'], model_setting['batch_size'],
use_weighted_sampler=False,
has_test_data=True,
is_training=False,
scaler=self.Scalers[ticker][model_type]['both'],
)
(y_pred_list, y_score_list) = get_predictions_both(test_loader, self.Models[ticker][model_type]['both'], device)
y_pred = y_pred_list[-1]
if model_type == 'hybrid':
col_feature_fundamental = model_setting["col_feature_fundamental_dict"][ticker]
col_feature_technical = model_setting["col_feature_technical_dict"][ticker]
test_df_windowed[col_target] = 1
ref_price = test_df_windowed[col_price].values[-1]
col_target_gains_thres = self.ModelParams[ticker][model_type]['col_target_gains_thres']
fundamental_mse = self.ModelParams[ticker][model_type]['fundamental_mse']
technical_mse = self.ModelParams[ticker][model_type]['technical_mse']
(_, _, test_loader, _) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_fundamental, col_target, None, None, test_df_windowed.copy(),
model_setting['window_size'], model_setting['batch_size'],
has_test_data=True,
is_training=False,
scaler=self.Scalers[ticker][model_type]['fundamental'],
)
y_pred_val = get_predictions_hybrid(
test_loader,
self.Models[ticker][model_type]['fundamental'],
self.Scalers[ticker][model_type]['fundamental'],
[col_price] + col_feature_fundamental,
device,
)
pred_price_fundamental = y_pred_val[-1]
pred_fundamental = get_regression_pred_decision(pred_price_fundamental - ref_price, col_target_gains_thres)
(_, _, test_loader, _) = get_torch_rnn_dataloaders_hybrid(
[col_price] + col_feature_technical, col_target, None, None, test_df_windowed.copy(),
model_setting['window_size'], model_setting['batch_size'],
has_test_data=True,
is_training=False,
scaler=self.Scalers[ticker][model_type]['technical'],
)
y_pred_val = get_predictions_hybrid(
test_loader,
self.Models[ticker][model_type]['technical'],
self.Scalers[ticker][model_type]['technical'],
[col_price] + col_feature_technical,
device,
)
pred_price_technical = y_pred_val[-1]
pred_technical = get_regression_pred_decision(pred_price_technical - ref_price, col_target_gains_thres)
y_pred = get_prediction_hybrid_regression(pred_fundamental, pred_technical, fundamental_mse, technical_mse)
for _signal in self.signals:
if ticker in self.signal_settings[_signal]['valid_tickers']:
pred_type = self.signal_settings[_signal]['pred_type']
lstm_ticker = self.signal_settings[_signal]['lstm_ticker']
if (pred_type == model_type) and (lstm_ticker == ticker):
self.prediction_dict[_signal][ticker] = y_pred
self.Signal[symbol][_signal].update_prediction_direction(self.prediction_dict[_signal][ticker])
# symbolQuantity = 0
if self.IsWarmingUp:
continue
# if self.Time.minute ==0 & :
# self.plot('rolling mdd','mdd', self.mdd)
# self.debug(self.mdd)
for _signal in self.signals:
if ticker in self.signal_settings[_signal]['valid_tickers']:
to_exit = self.Signal[symbol][_signal].check_exit(symbolData, data[symbol].Price, data[symbol].Time)
if to_exit:
self.Log(f"{str(self.Time)}: {ticker}: {_signal}: EXIT: {self.Signal[symbol][_signal].exitType}: Entry Price: {self.Signal[symbol][_signal].entryPrice:,.2f}: Exit Price: {data[symbol].Price:,.2f}: Quantity: {self.Signal[symbol][_signal].quantity}")
self.Signal[symbol][_signal].update_exit()
self.liquidate(symbol)
# elif (self.Time.weekday() == 4 and self.Time.hour == 16 and self.Time.minute == 30):
# # self.Log(f"{str(self.Time)}: {ticker}: {_signal}: EXIT: {self.Signal[symbol][_signal].exitType}: Entry Price: {self.Signal[symbol][_signal].entryPrice:,.2f}: Exit Price: {data[symbol].Price:,.2f}: Quantity: {self.Signal[symbol][_signal].quantity}")
# self.leftover_qty = self.Signal[symbol][_signal].quantity
# self.Signal[symbol][_signal].update_exit()
# self.liquidate(symbol)
# self.overweekend = 1
# self.leftover_ticker = symbol
self.UUP['trix'] = self.UUP__trix.current.value
self.UUP['macd'] = self.UUP__macd.current.value
self.UUP['roc'] = self.UUP__roc.current.value
has_enter = self.Signal[symbol][_signal].enter(symbolData, data[symbol].Price, self.portfolio[symbol].invested,data[symbol].Time, self.portfolio.total_portfolio_value, self.UUP)
if has_enter:
self.prev_equity = self.portfolio.total_portfolio_value
quantity = self.Signal[symbol][_signal].quantity * self.Signal[symbol][_signal].allocation_multiplier
quantity = int(np.ceil(quantity))
# self.Log(f"{str(self.Time)}: {ticker}: ENTRY Price: {data[symbol].Price:,.2f} | Quantity: {self.Signal[symbol][_signal].quantity} | Reason: {self.Signal[symbol][_signal].quote}")
price = self.Securities[symbol].Price
if quantity > 0:
# split the orders
self.entry_ticket = []
i = 0
while quantity > 10000000:
self.entry_ticket.append(self.LimitOrder(symbol, 10000000, round(price*(1+0.001),5)))
# self.Log(f"{str(self.Time)}: {ticker}: {_signal}: ENTRY: Price: {data[symbol].Price:,.2f}: Quantity: 10000000")
self.Log(f"{str(self.Time)}: {ticker}: ENTRY Price: {data[symbol].Price:,.2f} | Quantity: 10000000 | Reason: {self.Signal[symbol][_signal].quote}")
quantity -= 10000000
i += 1
self.entry_ticket.append(self.LimitOrder(symbol, quantity, round(price*(1+0.001),5)))
# self.Log(f"{str(self.Time)}: {ticker}: {_signal}: ENTRY: Price: {data[symbol].Price:,.2f}: Quantity: {quantity}")
self.Log(f"{str(self.Time)}: {ticker}: ENTRY Price: {data[symbol].Price:,.2f} | Quantity: {self.Signal[symbol][_signal].quantity} | Reason: {self.Signal[symbol][_signal].quote}")
elif quantity < 0:
self.entry_ticket = []
i = 0
while quantity < -10000000:
self.entry_ticket.append(self.LimitOrder(symbol, -10000000, round(price*(1-0.001),5)))
# self.Log(f"{str(self.Time)}: {ticker}: {_signal}: ENTRY: Price: {data[symbol].Price:,.2f}: Quantity: -10000000")
self.Log(f"{str(self.Time)}: {ticker}: ENTRY Price: {data[symbol].Price:,.2f} | Quantity: -10000000 | Reason: {self.Signal[symbol][_signal].quote}")
quantity += 10000000
i += 1
self.entry_ticket.append(self.LimitOrder(symbol, quantity, round(price*(1-0.001),5)))
# self.Log(f"{str(self.Time)}: {ticker}: {_signal}: ENTRY: Price: {data[symbol].Price:,.2f}: Quantity: {quantity}")
self.Log(f"{str(self.Time)}: {ticker}: ENTRY Price: {data[symbol].Price:,.2f} | Quantity: {self.Signal[symbol][_signal].quantity} | Reason: {self.Signal[symbol][_signal].quote}")
def on_margin_call(self, requests):
self.debug('Margin Call is coming')
self.Margin_Call = True
a = { "text": f"[US--{self.notification_value}--USDJPY Margin Call update]Margin Call is coming" }
payload = json.dumps(a)
self.notify.web("https://hooks.slack.com/services/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", payload)
return requests
def OnOrderEvent(self, orderEvent):
# self.Log(f'{orderEvent.OrderId}--{orderEvent.Status}--{orderEvent.quantity}')
if orderEvent.Status != OrderStatus.Filled:
return
# if self.CanTrade == False:
# self.Transactions.CancelOpenOrders()
IF_MARGIN = True
# self.debug(f"The order id is {orderEvent.OrderId}, and the message is {orderEvent.message}")
# Webhook Notification
price = orderEvent.FillPrice
quantity = orderEvent.quantity
self.debug(f"[US--{self.notification_value}--USDJPY order update] \nPrice: {price} \nQuantity: {quantity}")
a = { "text": f"[US--{self.notification_value}--USDJPY order update] \nPrice: {price} \nQuantity: {quantity}" }
payload = json.dumps(a)
self.notify.web("https://hooks.slack.com/services/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", payload)
if len(self.entry_ticket) > 0:
# When entry order is filled, place TP and SL orders
for i in range(len(self.entry_ticket)):
if (orderEvent.OrderId == self.entry_ticket[i].OrderId):
IF_MARGIN = False
price = orderEvent.FillPrice
quantity = orderEvent.absolute_fill_quantity
direction = orderEvent.direction
q1 = int (quantity/2)
q2 = quantity - q1
if direction == 0: # Long
# self.tp1_ticket.append((self.LimitOrder(orderEvent.Symbol, - q1, round(price*1.005,5))))
# self.tp2_ticket.append((self.LimitOrder(orderEvent.Symbol, - q2, round(price*1.0025,5))))
# self.sl_ticket.append((self.StopMarketOrder(orderEvent.Symbol, - quantity, round(price*0.997,5))))
self.tp1_ticket.append((self.LimitOrder(orderEvent.Symbol, - q1, round(price*1.008,5))))
self.tp2_ticket.append((self.LimitOrder(orderEvent.Symbol, - q2, round(price*1.004,5))))
self.sl_ticket.append((self.StopMarketOrder(orderEvent.Symbol, - quantity, round(price*0.993,5))))
# self.Log("TP/SL constructed")
if direction == 1: # Short
# self.tp1_ticket.append(self.LimitOrder(orderEvent.Symbol,q1, round(price*0.995,5)))
# self.tp2_ticket.append(self.LimitOrder(orderEvent.Symbol,q2, round(price*0.9975,5)))
# self.sl_ticket.append(self.StopMarketOrder(orderEvent.Symbol,quantity, round(price*1.0035,5)))
self.tp1_ticket.append(self.LimitOrder(orderEvent.Symbol,q1, round(price*0.992,5)))
self.tp2_ticket.append(self.LimitOrder(orderEvent.Symbol,q2, round(price*0.996,5)))
self.sl_ticket.append(self.StopMarketOrder(orderEvent.Symbol,quantity, round(price*1.003,5)))
# self.Log("TP/SL constructed")
#
return
if len(self.tp2_ticket) > 0:
for i in range(len(self.tp2_ticket)):
if (orderEvent.OrderId == self.tp2_ticket[i].OrderId):
IF_MARGIN = False
updateOrderFields = UpdateOrderFields()
updateOrderFields.Quantity = self.tp1_ticket[i].quantity
updateOrderFields.stop_price = self.entry_ticket[i].average_fill_price
self.sl_ticket[i].Update(updateOrderFields)
return
if len(self.tp1_ticket) > 0:
for i in range(len(self.tp1_ticket)):
if (orderEvent.OrderId == self.tp1_ticket[i].OrderId):
IF_MARGIN = False
self.sl_ticket[i].Cancel()
if len(self.entry_ticket)>i:
del self.entry_ticket[i]
del self.tp1_ticket[i]
del self.tp2_ticket[i]
del self.sl_ticket[i]
return
if len(self.sl_ticket) > 0:
for i in range(len(self.sl_ticket)):
if (orderEvent.OrderId == self.sl_ticket[i].OrderId):
IF_MARGIN = False
# self.Log(f'Stop Loss price is {orderEvent.FillPrice}')
self.tp1_ticket[i].Cancel()
self.tp2_ticket[i].Cancel()
if len(self.entry_ticket)>i:
del self.entry_ticket[i]
del self.sl_ticket[i]
del self.tp1_ticket[i]
del self.tp2_ticket[i]
# if self.portfolio['USDJPY'].invested:
# self.liquidate('USDJPY')
return
if self.Margin_Call:
i = 0
qty = orderEvent.quantity
self.Margin_Call = False
# self.debug(f'Hit margin call, the qty is {qty}')
while abs(qty) > 0:
if abs(qty) < abs(self.sl_ticket[i].quantity):
updateOrderFields = UpdateOrderFields()
updateOrderFields.Quantity = self.sl_ticket[i].quantity - qty
tmp_qty = self.sl_ticket[i].quantity - qty
self.sl_ticket[i].Update(updateOrderFields)
updateOrderFields = UpdateOrderFields()
updateOrderFields.Quantity = int(tmp_qty/2)
self.tp1_ticket[i].Update(updateOrderFields)
updateOrderFields = UpdateOrderFields()
updateOrderFields.Quantity = tmp_qty - int(tmp_qty/2)
self.tp2_ticket[i].Update(updateOrderFields)
return
elif abs(qty) == abs(self.sl_ticket[i].quantity):
self.sl_ticket[i].Cancel()
self.tp1_ticket[i].Cancel()
self.tp2_ticket[i].Cancel()
if len(self.entry_ticket)>i:
del self.entry_ticket[i]
del self.sl_ticket[i]
del self.tp1_ticket[i]
del self.tp2_ticket[i]
return
else:
self.sl_ticket[i].Cancel()
self.tp1_ticket[i].Cancel()
self.tp2_ticket[i].Cancel()
if len(self.entry_ticket)>i:
del self.entry_ticket[i]
del self.sl_ticket[i]
del self.tp1_ticket[i]
del self.tp2_ticket[i]
qty -= self.sl_ticket[i].quantity
i += 1
if self.CanTrade == False:
self.Transactions.CancelOpenOrders()
return
from AlgorithmImports import *
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader, WeightedRandomSampler, TensorDataset
import math
import random
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import confusion_matrix, classification_report
def get_class_distribution(obj):
count_dict = {
0: 0,
1: 0,
2: 0,
}
for i in obj:
count_dict[i] += 1
return count_dict
def get_weighted_sampler(y):
target_list = []
for t in y:
target_list.append(t)
target_list = torch.tensor(target_list)
class_count = [i for i in get_class_distribution(target_list.cpu().numpy()).values()]
class_weights = 1./torch.tensor(class_count, dtype=torch.float)
class_weights_all = class_weights[target_list]
weighted_sampler = WeightedRandomSampler(
weights=class_weights_all,
num_samples=len(class_weights_all),
replacement=True
)
return (weighted_sampler, class_weights)
def multi_acc(y_pred, y_test):
y_pred_softmax = torch.log_softmax(y_pred, dim = 1)
_, y_pred_tags = torch.max(y_pred_softmax, dim = 1)
correct_pred = (y_pred_tags == y_test).float()
acc = correct_pred.sum() / len(correct_pred)
acc = torch.round(acc * 100)
return acc
class EarlyStopping:
def __init__(self, tolerance=5, min_delta=0):
self.tolerance = tolerance
self.min_delta = min_delta
self.counter = 0
self.early_stop = False
def __call__(self, train_loss, validation_loss):
if (validation_loss - train_loss) > self.min_delta:
self.counter +=1
if self.counter >= self.tolerance:
self.early_stop = True
def get_rnn_dataloader_from_array(X_data, y_data, window_size, batch_size, is_test_loader=False, use_weighted_sampler=False):
weighted_sampler = None
class_weights = None
X, y = [], []
for i in range(window_size, len(X_data)+1):
feature = X_data[(i-window_size):i,:]
target = y_data[i-1]
X.append(feature)
y.append(target)
X = torch.tensor(X).float()
y = torch.tensor(y).long()
if is_test_loader:
data_loader = DataLoader(TensorDataset(X, y), batch_size=1)
else:
if use_weighted_sampler:
# (weighted_sampler, class_weights) = get_weighted_sampler(list(y_data), len(y))
(weighted_sampler, class_weights) = get_weighted_sampler(y)
data_loader = DataLoader(TensorDataset(X, y), sampler=weighted_sampler, batch_size=batch_size)
else:
data_loader = DataLoader(TensorDataset(X, y), shuffle=True, batch_size=batch_size)
return (data_loader, weighted_sampler, class_weights)
def get_torch_rnn_dataloaders(
col_feature, col_target, train_df, valid_df, test_df, window_size, batch_size,
use_weighted_sampler=False,
has_test_data=True,
is_training=True,
scaler=None,
):
if is_training:
X_train = train_df[col_feature].values
y_train = train_df[col_target].values
X_val = valid_df[col_feature].values
y_val = valid_df[col_target].values
if has_test_data:
X_test = test_df[col_feature].values
y_test = test_df[col_target].values
if is_training:
scaler = MinMaxScaler()
X_train = scaler.fit_transform(X_train)
X_val = scaler.transform(X_val)
if has_test_data:
X_test = scaler.transform(X_test)
if is_training:
X_train, y_train = np.array(X_train), np.array(y_train)
X_val, y_val = np.array(X_val), np.array(y_val)
if has_test_data:
X_test, y_test = np.array(X_test), np.array(y_test)
train_loader = None
val_loader = None
weighted_sampler = None
class_weights = None
if is_training:
(train_loader, weighted_sampler, class_weights) = get_rnn_dataloader_from_array(X_train, y_train, window_size, batch_size, is_test_loader=False, use_weighted_sampler=use_weighted_sampler)
(val_loader, _, _) = get_rnn_dataloader_from_array(X_val, y_val, window_size, batch_size, is_test_loader=False, use_weighted_sampler=False)
test_loader = None
if has_test_data:
(test_loader, _, _) = get_rnn_dataloader_from_array(X_test, y_test, window_size, batch_size, is_test_loader=True)
return (train_loader, val_loader, test_loader, scaler, weighted_sampler, class_weights)
def t2v(tau, f, out_features, w, b, w0, b0):
v1 = f(torch.matmul(tau, w) + b)
v2 = torch.matmul(tau, w0) + b0
return torch.cat([v1, v2], -1)
class SineActivation(nn.Module):
def __init__(self, in_features, out_features):
super(SineActivation, self).__init__()
self.out_features = out_features
self.w0 = nn.parameter.Parameter(torch.randn(in_features, 1))
self.b0 = nn.parameter.Parameter(torch.randn(1))
self.w = nn.parameter.Parameter(torch.randn(in_features, out_features-1))
self.b = nn.parameter.Parameter(torch.randn(out_features-1))
self.f = torch.sin
def forward(self, tau):
return t2v(tau, self.f, self.out_features, self.w, self.b, self.w0, self.b0)
class CosineActivation(nn.Module):
def __init__(self, in_features, out_features):
super(CosineActivation, self).__init__()
self.out_features = out_features
self.w0 = nn.parameter.Parameter(torch.randn(in_features, 1))
self.b0 = nn.parameter.Parameter(torch.randn(1))
self.w = nn.parameter.Parameter(torch.randn(in_features, out_features-1))
self.b = nn.parameter.Parameter(torch.randn(out_features-1))
self.f = torch.cos
def forward(self, tau):
return t2v(tau, self.f, self.out_features, self.w, self.b, self.w0, self.b0)
class GRUmodel(nn.Module):
def __init__(self, input_size, hidden_size, output_size, arc_num=1, use_t2v=True):
super(GRUmodel, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.output_size = output_size
self.arc_num = arc_num
self.use_t2v = use_t2v
if self.use_t2v:
self.t2v_layer = SineActivation(in_features=input_size, out_features=16)
self.layer0 = nn.Linear(16, input_size)
self.recurrent_layer = nn.GRU(input_size=input_size, hidden_size=hidden_size, num_layers=1, batch_first=True)
if self.arc_num == 1:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, output_size)
if self.arc_num == 2:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, output_size)
if self.arc_num == 3:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 64)
self.bn2 = nn.BatchNorm1d(64)
self.layer3 = nn.Linear(64, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, output_size)
def forward(self, x):
if len(x.shape) < 3:
x = x.unsqueeze(1)
if self.use_t2v:
x = self.t2v_layer(x)
x = self.layer0(x)
o, h = self.recurrent_layer(x)
h = h.squeeze().unsqueeze(0) if len(h.squeeze().shape) < 2 else h.squeeze()
if self.arc_num == 1:
x = self.layer1(h)
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
output = self.layer3(x)
if self.arc_num in [2,3]:
x = self.layer1(h)
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
x = self.layer3(x)
x = self.bn3(x)
output = self.layer4(x)
output if len(output.shape) > 1 else output.unsqueeze(0)
return output
class LSTMmodel(nn.Module):
def __init__(self, input_size, hidden_size, output_size,
use_dual_lstm=False, arc_num=0, use_t2v=True):
super(LSTMmodel, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.output_size = output_size
self.use_dual_lstm = use_dual_lstm
self.arc_num = arc_num
self.use_t2v = use_t2v
if self.use_t2v:
self.t2v_layer = SineActivation(in_features=input_size, out_features=16)
self.layer0 = nn.Linear(16, input_size)
self.recurrent_layer1 = nn.LSTM(input_size=input_size, hidden_size=hidden_size, num_layers=1, batch_first=True)
if self.use_dual_lstm:
self.recurrent_layer2 = nn.LSTM(input_size=hidden_size, hidden_size=hidden_size, num_layers=1, batch_first=True)
if self.arc_num == 0:
self.layer1 = nn.Linear(hidden_size, output_size)
if self.arc_num == 1:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, output_size)
if self.arc_num == 2:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, output_size)
if self.arc_num == 3:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 64)
self.bn2 = nn.BatchNorm1d(64)
self.layer3 = nn.Linear(64, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, output_size)
def forward(self, x):
if len(x.shape) < 3:
x = x.unsqueeze(1)
if self.use_t2v:
x = self.t2v_layer(x)
x = self.layer0(x)
rx, (hn, cn) = self.recurrent_layer1(x)
if self.use_dual_lstm:
rx, (hn, cn) = self.recurrent_layer2(rx)
if self.arc_num == 0:
output = self.layer1(rx[:,-1])
if self.arc_num == 1:
x = self.layer1(rx[:,-1])
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
output = self.layer3(x)
if self.arc_num in [2,3]:
x = self.layer1(rx[:,-1])
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
x = self.layer3(x)
x = self.bn3(x)
output = self.layer4(x)
output if len(output.shape) > 1 else output.unsqueeze(0)
return output
def get_rnn_model(
col_feature, train_loader, val_loader,
epochs, batch_size, learning_rate, window_size, hidden_size, device,
use_early_stop=False, use_weighted_sampler=False, class_weights=None,
use_dual_lstm=False, use_gru_model=False,
):
input_size = len(col_feature)
output_size = 3
#use_early_stop = False
if use_weighted_sampler:
criterion = nn.CrossEntropyLoss(weight=class_weights.to(device))
else:
criterion = nn.CrossEntropyLoss()
# encoder = RNNEncoder(input_size, hidden_size, device).to(device)
# decoder = RNNDecoder(hidden_size, output_size).to(device)
# model = RNNSeq2Seq(encoder, decoder).to(device)
if use_gru_model:
model = GRUmodel(input_size, hidden_size, output_size).to(device)
else:
model = LSTMmodel(input_size, hidden_size, output_size, use_dual_lstm=use_dual_lstm).to(device)
optimizer = optim.Adam(model.parameters(), lr=learning_rate)
accuracy_stats = {
'train': [],
"val": []
}
loss_stats = {
'train': [],
"val": []
}
if use_early_stop:
early_stopping = EarlyStopping(tolerance=5, min_delta=0.01)
# print("Begin training.")
for e in range(1, epochs+1):
# TRAINING
train_epoch_loss = 0
train_epoch_acc = 0
model.train()
for X_train_batch, y_train_batch in train_loader:
if X_train_batch.shape[0] == 1:
continue
X_train_batch, y_train_batch = X_train_batch.to(device), y_train_batch.to(device)
optimizer.zero_grad()
y_train_pred = model(X_train_batch)
train_loss = criterion(y_train_pred, y_train_batch)
train_acc = multi_acc(y_train_pred, y_train_batch)
train_loss.backward()
optimizer.step()
train_epoch_loss += train_loss.item()
train_epoch_acc += train_acc.item()
# VALIDATION
model.eval()
with torch.no_grad():
val_epoch_loss = 0
val_epoch_acc = 0
for X_val_batch, y_val_batch in val_loader:
X_val_batch, y_val_batch = X_val_batch.to(device), y_val_batch.to(device)
y_val_pred = model(X_val_batch)
val_loss = criterion(y_val_pred, y_val_batch)
val_acc = multi_acc(y_val_pred, y_val_batch)
val_epoch_loss += val_loss.item()
val_epoch_acc += val_acc.item()
loss_stats['train'].append(train_epoch_loss/len(train_loader))
loss_stats['val'].append(val_epoch_loss/len(val_loader))
accuracy_stats['train'].append(train_epoch_acc/len(train_loader))
accuracy_stats['val'].append(val_epoch_acc/len(val_loader))
if use_early_stop:
early_stopping(train_epoch_loss/len(train_loader), val_epoch_loss/len(val_loader))
# print(f'Epoch {e+0:03}: | Train Loss: {train_epoch_loss/len(train_loader):.5f} | Val Loss: {val_epoch_loss/len(val_loader):.5f} | Train Acc: {train_epoch_acc/len(train_loader):.3f}| Val Acc: {val_epoch_acc/len(val_loader):.3f}')
if use_early_stop and early_stopping.early_stop:
break
return model
def get_predictions(test_loader, model, device):
y_pred_list = []
y_score_list = []
with torch.no_grad():
model.eval()
for X_batch, _ in test_loader:
X_batch = X_batch.to(device)
y_test_pred = model(X_batch)
y_pred_score, y_pred_tags = torch.max(y_test_pred, dim = 1)
y_score_list.append(y_pred_score.cpu().numpy())
y_pred_list.append(y_pred_tags.cpu().numpy())
y_pred_list = [a.squeeze().tolist() for a in y_pred_list]
y_score_list = [a.squeeze().tolist() for a in y_score_list]
return (y_pred_list, y_score_list)
def get_prediction_hybrid(row, original=False):
out = None
if original:
if (row['pred_technical'] == 1):
out = 1
elif (row['pred_fundamental'] == 1):
out = 1
elif row['pred_technical'] == row['pred_fundamental']:
out = row['pred_technical']
else:
if row['score_technical'] >= row['score_fundamental']:
out = row['pred_technical']
else:
out = row['pred_fundamental']
else:
if row['pred_technical'] == row['pred_fundamental']:
out = row['pred_technical']
else:
out = 1
return out
def get_prediction_hybrid_max(row):
out = None
if row['score_technical'] >= row['score_fundamental']:
out = row['pred_technical']
else:
out = row['pred_fundamental']
return out
def get_prediction_hybrid_greedy(row):
out = None
if row['pred_technical'] == row['pred_fundamental']:
out = row['pred_technical']
elif row['pred_technical'] == 1:
out = row['pred_fundamental']
elif row['pred_fundamental'] == 1:
out = row['pred_technical']
else:
if row['score_technical'] >= row['score_fundamental']:
out = row['pred_technical']
else:
out = row['pred_fundamental']
return out
#region imports
from AlgorithmImports import *
#endregion
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader, WeightedRandomSampler, TensorDataset
from sklearn.preprocessing import MinMaxScaler
import torch.nn.functional as F
from sklearn.metrics import mean_squared_error
import math
import random
import pandas as pd
import numpy as np
def get_rnn_dataloader_from_array(X_data, y_data, window_size, batch_size, is_test_loader=False):
X, y = [], []
for i in range(window_size, len(X_data)+1):
feature = X_data[(i-window_size):i,:]
target = y_data[i-1]
X.append(feature)
y.append(target)
X = torch.tensor(X).float()
# y = torch.tensor(y).long()
y = torch.tensor(y).float()
if is_test_loader:
data_loader = DataLoader(TensorDataset(X, y), batch_size=1)
else:
data_loader = DataLoader(TensorDataset(X, y), shuffle=True, batch_size=batch_size)
return data_loader
def get_torch_rnn_dataloaders(
col_feature, col_target, train_df, valid_df, test_df, window_size, batch_size,
has_test_data=True,
is_training=True,
scaler=None,
):
if is_training:
train_data = train_df[[col_target]+col_feature].values
valid_df_windowed = pd.concat([train_df,valid_df]).copy()
valid_df_windowed = valid_df_windowed.tail(len(valid_df) + window_size-1)
valid_data = valid_df_windowed[[col_target]+col_feature].values
if has_test_data:
test_data = test_df[[col_target]+col_feature].values
if is_training:
scaler = MinMaxScaler(feature_range=(0, 1))
train_data = scaler.fit_transform(train_data)
valid_data = scaler.transform(valid_data)
if has_test_data:
test_data = scaler.transform(test_data)
if is_training:
X_train = train_data[:, 1:]
y_train = train_data[:, 0]
X_val = valid_data[:, 1:]
y_val = valid_data[:, 0]
if has_test_data:
X_test = test_data[:, 1:]
y_test = test_data[:, 0]
train_loader = None
val_loader = None
test_loader = None
if is_training:
train_loader = get_rnn_dataloader_from_array(X_train, y_train, window_size, batch_size, is_test_loader=False)
val_loader = get_rnn_dataloader_from_array(X_val, y_val, window_size, batch_size, is_test_loader=True)
if has_test_data:
test_loader = get_rnn_dataloader_from_array(X_test, y_test, window_size, batch_size, is_test_loader=True)
return (train_loader, val_loader, test_loader, scaler)
def t2v(tau, f, out_features, w, b, w0, b0):
v1 = f(torch.matmul(tau, w) + b)
v2 = torch.matmul(tau, w0) + b0
return torch.cat([v1, v2], -1)
class SineActivation(nn.Module):
def __init__(self, in_features, out_features):
super(SineActivation, self).__init__()
self.out_features = out_features
self.w0 = nn.parameter.Parameter(torch.randn(in_features, 1))
self.b0 = nn.parameter.Parameter(torch.randn(1))
self.w = nn.parameter.Parameter(torch.randn(in_features, out_features-1))
self.b = nn.parameter.Parameter(torch.randn(out_features-1))
self.f = torch.sin
def forward(self, tau):
return t2v(tau, self.f, self.out_features, self.w, self.b, self.w0, self.b0)
class CosineActivation(nn.Module):
def __init__(self, in_features, out_features):
super(CosineActivation, self).__init__()
self.out_features = out_features
self.w0 = nn.parameter.Parameter(torch.randn(in_features, 1))
self.b0 = nn.parameter.Parameter(torch.randn(1))
self.w = nn.parameter.Parameter(torch.randn(in_features, out_features-1))
self.b = nn.parameter.Parameter(torch.randn(out_features-1))
self.f = torch.cos
def forward(self, tau):
return t2v(tau, self.f, self.out_features, self.w, self.b, self.w0, self.b0)
class GRUmodel(nn.Module):
def __init__(self, input_size, hidden_size, arc_num=1, use_t2v=True):
super(GRUmodel, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.arc_num = arc_num
self.use_t2v = use_t2v
if self.use_t2v:
self.t2v_layer = SineActivation(in_features=input_size, out_features=16)
self.layer0 = nn.Linear(16, input_size)
self.recurrent_layer = nn.GRU(input_size=input_size, hidden_size=hidden_size, num_layers=1, batch_first=True)
if self.arc_num == 1:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, 1)
if self.arc_num == 2:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, 1)
if self.arc_num == 3:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 64)
self.bn2 = nn.BatchNorm1d(64)
self.layer3 = nn.Linear(64, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, 1)
def forward(self, x):
if len(x.shape) < 3:
x = x.unsqueeze(1)
if self.use_t2v:
x = self.t2v_layer(x)
x = self.layer0(x)
o, h = self.recurrent_layer(x)
h = h.squeeze().unsqueeze(0) if len(h.squeeze().shape) < 2 else h.squeeze()
if self.arc_num == 1:
x = self.layer1(h)
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
output = self.layer3(x)
if self.arc_num in [2,3]:
x = self.layer1(h)
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
x = self.layer3(x)
x = self.bn3(x)
output = self.layer4(x)
output if len(output.shape) > 1 else output.unsqueeze(0)
return output
class LSTMmodel(nn.Module):
def __init__(self, input_size, hidden_size,
use_dual_lstm=False, arc_num=1, use_t2v=True):
super(LSTMmodel, self).__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.use_dual_lstm = use_dual_lstm
self.arc_num = arc_num
self.use_t2v = use_t2v
if self.use_t2v:
self.t2v_layer = SineActivation(in_features=input_size, out_features=16)
self.layer0 = nn.Linear(16, input_size)
self.recurrent_layer1 = nn.LSTM(input_size=input_size, hidden_size=hidden_size, num_layers=1, batch_first=True)
if self.use_dual_lstm:
self.recurrent_layer2 = nn.LSTM(input_size=hidden_size, hidden_size=hidden_size, num_layers=1, batch_first=True)
if self.arc_num == 0:
self.layer1 = nn.Linear(hidden_size, 1)
if self.arc_num == 1:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, 1)
if self.arc_num == 2:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 256)
self.bn2 = nn.BatchNorm1d(256)
self.layer3 = nn.Linear(256, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, 1)
if self.arc_num == 3:
self.layer1 = nn.Linear(hidden_size, 128)
self.bn1 = nn.BatchNorm1d(128)
self.layer2 = nn.Linear(128, 64)
self.bn2 = nn.BatchNorm1d(64)
self.layer3 = nn.Linear(64, 32)
self.bn3 = nn.BatchNorm1d(32)
self.layer4 = nn.Linear(32, 1)
def forward(self, x):
if len(x.shape) < 3:
x = x.unsqueeze(1)
if self.use_t2v:
x = self.t2v_layer(x)
x = self.layer0(x)
rx, (hn, cn) = self.recurrent_layer1(x)
if self.use_dual_lstm:
rx, (hn, cn) = self.recurrent_layer2(rx)
if self.arc_num == 0:
output = self.layer1(rx[:,-1])
if self.arc_num == 1:
x = self.layer1(rx[:,-1])
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
output = self.layer3(x)
if self.arc_num in [2,3]:
x = self.layer1(rx[:,-1])
x = self.bn1(x)
x = self.layer2(x)
x = self.bn2(x)
x = self.layer3(x)
x = self.bn3(x)
output = self.layer4(x)
output if len(output.shape) > 1 else output.unsqueeze(0)
return output
def get_rnn_model(
col_feature, train_loader, val_loader,
epochs, learning_rate, hidden_size, device,
use_dual_lstm=False, use_gru_model=False,
):
input_size = len(col_feature)
if use_gru_model:
model = GRUmodel(input_size, hidden_size).to(device)
else:
model = LSTMmodel(input_size, hidden_size, use_dual_lstm=use_dual_lstm).to(device)
optimizer = optim.Adam(model.parameters(), lr=learning_rate)
loss_stats = {
'train': [],
"val": []
}
for e in range(1, epochs+1):
# TRAINING
train_epoch_loss = 0
model.train()
for X_train_batch, y_train_batch in train_loader:
if X_train_batch.shape[0] == 1:
continue
X_train_batch, y_train_batch = X_train_batch.to(device), y_train_batch.to(device)
optimizer.zero_grad()
y_train_pred = model(X_train_batch)
train_loss = F.smooth_l1_loss(y_train_pred, y_train_batch.unsqueeze(1))
train_loss.backward()
optimizer.step()
train_epoch_loss += train_loss.item()
# VALIDATION
model.eval()
with torch.no_grad():
val_epoch_loss = 0
for X_val_batch, y_val_batch in val_loader:
X_val_batch, y_val_batch = X_val_batch.to(device), y_val_batch.to(device)
y_val_pred = model(X_val_batch)
val_loss = F.smooth_l1_loss(y_val_pred, y_val_batch.unsqueeze(1))
val_epoch_loss += val_loss.item()
loss_stats['train'].append(train_epoch_loss/len(train_loader))
loss_stats['val'].append(val_epoch_loss/len(val_loader))
return model
def get_predictions(test_loader, model, scaler, col_feature, device):
y_pred_list = []
with torch.no_grad():
model.eval()
for X_batch, _ in test_loader:
X_batch = X_batch.to(device)
y_test_pred = model(X_batch)
y_test_pred = y_test_pred.cpu().squeeze().numpy().item()
y_pred_list.append(y_test_pred)
def inverse_transform(y_pred, col_feature):
extended = np.zeros((len(y_pred), len(col_feature)+1))
extended[:, 0] = y_pred
return scaler.inverse_transform(extended)[:, 0]
y_pred = np.array(y_pred_list)
y_pred = inverse_transform(y_pred, col_feature)
return y_pred
def get_prediction_hybrid_regression(pred_fundamental, pred_technical, fundamental_mse, technical_mse):
out = None
# if (pred_technical == 1):
# out = 1
# elif (pred_fundamental == 1):
# out = 1
if pred_technical == pred_fundamental:
out = pred_technical
else:
if technical_mse <= fundamental_mse:
out = pred_technical
else:
out = 1
return pred_technical
def get_regression_pred_decision(diff, col_target_gains_thres):
if diff > col_target_gains_thres:
return 2
if -diff > col_target_gains_thres:
return 0
else:
return 1
from AlgorithmImports import *
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader, WeightedRandomSampler, TensorDataset
import math
import random
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import confusion_matrix, classification_report
def set_seed(seed: int = 100) -> None:
np.random.seed(seed)
random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
# When running on the CuDNN backend, two further options must be set
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# Set a fixed value for the hash seed
os.environ["PYTHONHASHSEED"] = str(seed)
# print(f"Random seed set as {seed}")
##-##
# def get_upper_threshold(close):
# difference = close.diff()
# difference[0] = 0
# difference = difference.abs()
# bins = pd.cut(difference, bins=10)
# bins = bins.value_counts().to_frame().reset_index()
# bins["index"] = bins["index"].apply(lambda x: x.right)
# bins = bins.to_numpy()
# percentile_count = len(difference) * 0.85
# count = 0
# for i in range(10):
# count += bins[i, 1]
# if count > percentile_count:
# return bins[i, 0]
def get_upper_threshold(close):
difference = close.diff()
difference[0] = 0
difference = difference.abs()
bins = pd.cut(difference, bins=10)
bins = bins.value_counts().to_frame().reset_index()
bins.columns = ['close','count']
bins["close"] = bins["close"].apply(lambda x: x.right)
bins = bins.to_numpy()
percentile_count = len(difference) * 0.85
count = 0
for i in range(10):
count += bins[i, 1]
if count > percentile_count:
return bins[i, 0]
##-##
def get_entropy(labels, base=None):
vc = pd.Series(labels).value_counts(normalize=True, sort=False)
base = math.e if base is None else base
return -(vc * np.log(vc)/np.log(base)).sum()
def get_threshold(close):
difference = close.diff()
difference = difference.drop(0)
difference = difference.tolist()
threshold = 0
thres_upper_bound = get_upper_threshold(close)
temp_thres = 0
best_entropy = -float('inf')
while temp_thres < thres_upper_bound:
labels = []
for diff in difference:
if diff > temp_thres:
labels.append(2)
elif -diff > temp_thres:
labels.append(1)
else:
labels.append(0)
entropy = get_entropy(labels)
if entropy > best_entropy:
best_entropy = entropy
threshold = temp_thres
temp_thres = temp_thres + 0.00001
return np.round(threshold,5)
from AlgorithmImports import *
import pandas as pd
import numpy as np
from datetime import datetime, timedelta
from collections import deque
from utils import (
getFxPositionSize,
)
class FxLstmSignal:
def __init__(
self,
algorithm,
symbol,
ticker,
general_setting,
signal_setting,
):
# General Initializations
self.algorithm = algorithm
IS_PAPER = self.algorithm.get_parameter("IS_PAPER")
if IS_PAPER == 'True':
self.notification_value = 'PAPER'
else:
self.notification_value = 'REAL'
self.symbol = symbol
self.ticker = ticker
self.general_setting = general_setting
self.signal_setting = signal_setting
self.prediction_direction_map_dict = self.signal_setting['prediction_direction_map_dict']
self.valid_tickers = signal_setting["valid_tickers"]
self.enter_long_trades = signal_setting["enter_long_trades"]
self.enter_short_trades = signal_setting["enter_short_trades"]
self.prediction_direction = 0
self.use_exit_reason = self.signal_setting["useTralingStop"]
self.direction = 0
self.quantity = 0
self.entryBarCount = 0
self.allocation_multiplier = 1
self.entryPrice = None
self.stopPrice = None
self.stopDistance = None
self.targetPrice = None
self.targetDistance = None
self.inTrade = False
self.lookForExit = False
self.trailingStop = False
self.exitType = None
self.macd = None
self.quote = None
self.max_qty = None
def update_prediction_direction(self, prediction):
self.prediction_direction = self.prediction_direction_map_dict[prediction]
def enter(self, symbolData, price, position, market_time, total_port_value, uup):
# max_qty = int(total_port_value/(0.02*price))-1000
# max_qty = min(max_qty, 30000000)
# self.max_qty = max_qty
margin_qty = int(total_port_value/(0.05*(1/price)))-1000 # Margin ratio = 20:1; 1/20 =0.05
# margin_qty = 30000
limit_qty = 30000000
max_qty = min(margin_qty, limit_qty)
self.max_qty = max_qty
has_enter = False
self.inTrade = position
if self.inTrade :
active_consolidator = symbolData.consolidators[self.signal_setting["active_timeframe"]]
ATR = active_consolidator.indicators[f"ATR{self.signal_setting['atrLength']}"]
SMA = active_consolidator.indicators[f"SMA{self.signal_setting['sma_filter_lookback']}"]
AO = active_consolidator.indicators[f"AO"]
ROC = active_consolidator.indicators[f"ROC{self.signal_setting['roc_filter_lookback']}"]
TRIX = active_consolidator.indicators[f"TRIX{self.signal_setting['trix_filter_lookback']}"]
MACD = active_consolidator.indicators[f"MACD"]
MACD1 = active_consolidator.indicators[f"MACD"]
self.macd = MACD["macd"][0]
if self.algorithm.Time.minute == 0:
self.algorithm.Log(f"{self.notification_value} MACD is {MACD['macd'][0]} ; AO is {AO['val'][0]}; Prediction is {self.prediction_direction}; UUP_roc is {round(uup['roc'],5)} ")
a = { "text": f"[US--{self.notification_value}--USDJPY hourly update] MACD is {MACD['macd'][0]} ; AO is {AO['val'][0]}; Prediction is {self.prediction_direction}; UUP_roc is {round(uup['roc'],5)}" }
payload = json.dumps(a)
self.algorithm.notify.web("https://hooks.slack.com/services/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", payload)
return False
active_consolidator = symbolData.consolidators[self.signal_setting["active_timeframe"]]
ATR = active_consolidator.indicators[f"ATR{self.signal_setting['atrLength']}"]
SMA = active_consolidator.indicators[f"SMA{self.signal_setting['sma_filter_lookback']}"]
AO = active_consolidator.indicators[f"AO"]
ROC = active_consolidator.indicators[f"ROC{self.signal_setting['roc_filter_lookback']}"]
TRIX = active_consolidator.indicators[f"TRIX{self.signal_setting['trix_filter_lookback']}"]
MACD = active_consolidator.indicators[f"MACD"]
MACD1 = active_consolidator.indicators[f"MACD"]
self.macd = MACD1["macd"][0]
# SMA_fast = active_consolidator.indicators[f"SMA{self.signal_setting['sma_filter_lookback_fast']}"]
# SMA_slow = active_consolidator.indicators[f"SMA{self.signal_setting['sma_filter_lookback_slow']}"]
if self.algorithm.Time.minute == 0:
self.algorithm.Log(f"{self.notification_value} TRIX is {TRIX['val'][0]}; Prediction is {self.prediction_direction}; UUP_roc is {round(uup['roc'],5)} ")
a = { "text": f"[US--{self.notification_value}--USDJPY hourly update] TRIX is {TRIX['val'][0]}; Prediction is {self.prediction_direction}; UUP_roc is {round(uup['roc'],5)} " }
payload = json.dumps(a)
self.algorithm.notify.web("https://hooks.slack.com/services/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", payload)
if self.signal_setting["use_sma_filter"]:
is_long_trend_sma = price < SMA["val"][0]
is_short_trend_sma = price > SMA["val"][0]
else:
is_long_trend_sma = False
is_short_trend_sma = False
if self.signal_setting["use_ao_filter"]:
is_long_trend_ao = (AO["val"][0] > 0)
is_short_trend_ao = (AO["val"][0] < 0)
else:
is_long_trend_ao = False
is_short_trend_ao = False
if self.signal_setting["use_roc_filter"]:
is_long_trend_roc = (ROC["val"][0] < 0)
is_short_trend_roc = (ROC["val"][0] > 0 )
else:
is_long_trend_roc = False
is_short_trend_roc = False
if self.signal_setting["use_trix_filter"]:
is_long_trend_trix = (TRIX["val"][0] < 0 )
is_short_trend_trix = (TRIX["val"][0] > 0 )
else:
is_long_trend_trix = False
is_short_trend_trix = False
if self.signal_setting["use_macd_filter"]:
is_long_trend_macd = (MACD["macd"][0] < 0 )
is_short_trend_macd = (MACD["macd"][0] > 0 )
else:
is_long_trend_macd = False
is_short_trend_macd = False
if is_short_trend_trix and (self.prediction_direction < 0) and uup['roc'] >0 :
has_enter = True
self.inTrade = True
self.direction = 1
if self.signal_setting["use_movement_thres_for_stops"]:
stopSize = self.signal_setting["movement_thres"] * self.signal_setting["longStopMultiplier"]
longStopPrice = active_consolidator.close[0] - stopSize
longStopDistance = active_consolidator.close[0] - longStopPrice
else:
stopSize = ATR["val"][0] * self.signal_setting["longStopMultiplier"]
longPriceSource = min(active_consolidator.low[0], active_consolidator.low[1])
longStopPrice = longPriceSource - stopSize
longStopDistance = active_consolidator.close[0] - longStopPrice
self.targetPrice = active_consolidator.close[0] + longStopDistance * self.signal_setting["longRiskRewardMultiplier"]
self.targetDistance = self.targetPrice - active_consolidator.close[0]
self.quantity = self.direction * getFxPositionSize(
longStopDistance,
self.signal_setting["risk_pct"],
self.algorithm,
self.symbol,
)
self.algorithm.Log(f"Long position: TRIX is {TRIX['val'][0]} ; prediction is {self.prediction_direction} ")
self.quote = 'Long TRIX & prediction'
if abs(self.quantity) > max_qty:
self.quantity = self.direction * max_qty
self.entryPrice = price
self.stopPrice = longStopPrice
self.stopDistance = longStopDistance
self.entryBarCount = active_consolidator.BarCount
self.trailingStop = False
self.lookForExit = False
elif is_short_trend_macd and is_short_trend_sma and uup['roc'] >0:
has_enter = True
self.inTrade = True
self.direction = 1
if self.signal_setting["use_movement_thres_for_stops"]:
stopSize = self.signal_setting["movement_thres"] * self.signal_setting["longStopMultiplier"]
longStopPrice = active_consolidator.close[0] - stopSize
longStopDistance = active_consolidator.close[0] - longStopPrice
else:
stopSize = ATR["val"][0] * self.signal_setting["longStopMultiplier"]
longPriceSource = min(active_consolidator.low[0], active_consolidator.low[1])
longStopPrice = longPriceSource - stopSize
longStopDistance = active_consolidator.close[0] - longStopPrice
self.targetPrice = active_consolidator.close[0] + longStopDistance * self.signal_setting["longRiskRewardMultiplier"]
self.targetDistance = self.targetPrice - active_consolidator.close[0]
self.quantity = self.direction * getFxPositionSize(
longStopDistance,
self.signal_setting["risk_pct"],
self.algorithm,
self.symbol,
)
self.algorithm.Log(f"Long position: MACD is {MACD['macd'][0]} ; SMA is {SMA['val'][0]} ")
self.quote = 'Long MACD & prediction'
if abs(self.quantity) > max_qty:
self.quantity = self.direction * max_qty
self.entryPrice = price
self.stopPrice = longStopPrice
self.stopDistance = longStopDistance
self.entryBarCount = active_consolidator.BarCount
self.trailingStop = False
self.lookForExit = False
elif is_long_trend_trix and (self.prediction_direction > 0) and uup['roc'] < 0:
has_enter = True
self.inTrade = True
self.direction = -1
if self.signal_setting["use_movement_thres_for_stops"]:
stopSize = self.signal_setting["movement_thres"] * self.signal_setting["shortStopMultiplier"]
shortStopPrice = active_consolidator.close[0] + stopSize
shortStopDistance = shortStopPrice - active_consolidator.close[0]
else:
stopSize = ATR["val"][0] * self.signal_setting["shortStopMultiplier"]
shortPriceSource = max(active_consolidator.high[0], active_consolidator.high[1])
shortStopPrice = shortPriceSource + stopSize
shortStopDistance = shortStopPrice - active_consolidator.close[0]
self.targetPrice = active_consolidator.close[0] - shortStopDistance * self.signal_setting["shortRiskRewardMultiplier"]
self.targetDistance = active_consolidator.close[0] - self.targetPrice
self.quantity = self.direction * getFxPositionSize(
shortStopDistance,
self.signal_setting["risk_pct"],
self.algorithm,
self.symbol,
)
self.algorithm.Log(f"Short position: MACD is {MACD['macd'][0]} ; prediction is {self.prediction_direction} ")
self.quote = 'Short MACD & prediction'
if abs(self.quantity) > max_qty:
self.quantity = self.direction * max_qty
self.entryPrice = price
self.stopPrice = shortStopPrice
self.stopDistance = shortStopDistance
self.entryBarCount = active_consolidator.BarCount
self.trailingStop = False
self.lookForExit = False
return has_enter
def check_exit(self, symbolData, price, market_time):
if not self.inTrade:
active_consolidator = symbolData.consolidators[self.signal_setting["active_timeframe"]]
ATR = active_consolidator.indicators[f"ATR{self.signal_setting['atrLength']}"]
SMA = active_consolidator.indicators[f"SMA{self.signal_setting['sma_filter_lookback']}"]
AO = active_consolidator.indicators[f"AO"]
ROC = active_consolidator.indicators[f"ROC{self.signal_setting['roc_filter_lookback']}"]
TRIX = active_consolidator.indicators[f"TRIX{self.signal_setting['trix_filter_lookback']}"]
MACD = active_consolidator.indicators[f"MACD"]
MACD1 = active_consolidator.indicators[f"MACD"]
MACD = active_consolidator.indicators[f"MACD"]
self.macd = MACD["macd"][0]
return False
to_exit = False
active_consolidator = symbolData.consolidators[self.signal_setting["active_timeframe"]]
can_exit = (active_consolidator.BarCount - self.entryBarCount) > self.signal_setting["exit_wait_period"]
ATR = active_consolidator.indicators[f"ATR{self.signal_setting['atrLength']}"]
SMA = active_consolidator.indicators[f"SMA{self.signal_setting['sma_filter_lookback']}"]
AO = active_consolidator.indicators[f"AO"]
ROC = active_consolidator.indicators[f"ROC{self.signal_setting['roc_filter_lookback']}"]
TRIX = active_consolidator.indicators[f"TRIX{self.signal_setting['trix_filter_lookback']}"]
MACD1 = active_consolidator.indicators[f"MACD"]
MACD = active_consolidator.indicators[f"MACD"]
self.macd = MACD["macd"][0]
if (self.direction > 0):
if self.signal_setting["use_trix_for_stops"]:
if (TRIX["val"][0] > 0 ):
to_exit = True
self.exitType = 'DirectionChange'
if self.signal_setting["use_sma_for_stops"] :
if (SMA["val"][0] > price ) & (self.prediction_direction < 0):
to_exit = True
self.exitType = 'DirectionChange'
if self.signal_setting["use_roc_for_stops"]:
if (ROC["val"][0] > 0 ):
to_exit = True
self.exitType = 'DirectionChange'
if self.signal_setting["use_macd_for_stops"]:
if self.signal_setting["use_ao_for_stops"]:
if (MACD["macd"][0] > 0 and AO["val"][0] < 0 ):
to_exit = True
self.exitType = 'DirectionChange'
if (self.direction < 0):
if self.signal_setting["use_trix_for_stops"]:
if (TRIX["val"][0] < 0 ):
to_exit = True
self.exitType = 'DirectionChange'
if self.signal_setting["use_sma_for_stops"] :
if (SMA["val"][0] < price )& (self.prediction_direction > 0):
to_exit = True
self.exitType = 'DirectionChange'
if self.signal_setting["use_roc_for_stops"]:
if (ROC["val"][0] < 0 ):
to_exit = True
self.exitType = 'DirectionChange'
if self.signal_setting["use_macd_for_stops"]:
if self.signal_setting["use_ao_for_stops"]:
if (MACD["macd"][0] < 0 and AO["val"][0] > 0 ):
to_exit = True
self.exitType = 'DirectionChange'
if to_exit and can_exit:
return to_exit
if (self.direction > 0):
if (price >= self.targetPrice) and self.use_exit_reason:
self.lookForExit = True
if (self.direction > 0):
if self.signal_setting["useTralingStop"] and self.lookForExit:
trail = active_consolidator.close[0] - ATR["val"][0] * self.signal_setting["trailStopSize"]
if trail > self.stopPrice:
self.stopPrice = trail
self.trailingStop = True
if (self.direction > 0):
if (self.signal_setting["useTralingStop"] and self.lookForExit and (price <= self.stopPrice)):
to_exit = True
self.exitType = 'TrailingStopLoss'
if (self.direction < 0):
if (price <= self.targetPrice) and self.use_exit_reason:
self.lookForExit = True
if (self.direction < 0):
if self.signal_setting["useTralingStop"] and self.lookForExit:
trail = active_consolidator.close[0] + ATR["val"][0] * self.signal_setting["trailStopSize"]
if trail < self.stopPrice:
self.stopPrice = trail
self.trailingStop = True
if (self.direction < 0):
if (self.signal_setting["useTralingStop"] and self.lookForExit and (price >= self.stopPrice)):
to_exit = True
self.exitType = 'TrailingStopLoss'
to_exit = to_exit and can_exit
return to_exit
def update_exit(self):
self.direction = 0
self.quantity = 0
self.entryBarCount = 0
self.allocation_multiplier = 1
self.entryPrice = None
self.stopPrice = None
self.stopDistance = None
self.targetPrice = None
self.targetDistance = None
self.inTrade = False
self.lookForExit = False
self.trailingStop = False
self.exitType = None
from AlgorithmImports import *
import scipy
import math
def getFxPositionSize(stopPoints, riskPct, algorithm, symbol):
Balance = algorithm.Portfolio.TotalPortfolioValue
LotSize = algorithm.Securities[symbol].SymbolProperties.LotSize
price = algorithm.Securities[symbol].Close
conversionRate = algorithm.Securities[symbol].QuoteCurrency.ConversionRate
pointValue = LotSize * conversionRate
# pointValue = 0.0001*Balance/price
units = int(np.ceil((Balance * riskPct) / (stopPoints * pointValue)))
units = int(np.ceil(10000*(Balance * riskPct) / ((stopPoints/0.0001) * pointValue)))
# units = 1000*int(np.ceil((Balance * riskPct) / ( 0.5 * pointValue ))) # default set stopPoints as 80, abandon longstopdistance
# units = int(Balance * riskPct / price)
# units = 10000*int(np.ceil((Balance * riskPct) / (40 * pointValue)))
return units