refactor(Project): move out load_data to utils, rename fetch_data to run_fetch_data, got classifiers to work (#38)

.github/workflows/test.yml

@@ -21,6 +21,10 @@ jobs:
 #           WANDB_API_KEY: ${{ secrets.WANDB_API_KEY }}
         run: |
           pytest --junit-xml pytest.xml
+      - name: Run pipeline
+        shell: bash -l {0}
+        run: |
+          python run_pipeline.py
       - name: Upload Unit Test Results
         if: always()
         uses: actions/upload-artifact@v2

archive/forecasting.py

@@ -9,7 +9,7 @@ from pytorch_forecasting.metrics import QuantileLoss
 import sys
 sys.path.insert(0, '..')
 
-from load_data import load_files
+from utils.load_data import load_files
 
 print("success")
 

archive/lightning/create_dataset.py

@@ -8,7 +8,7 @@ from pytorch_forecasting.metrics import QuantileLoss
 import sys
 sys.path.insert(0, '..')
 
-from load_data import load_files
+from utils.load_data import load_files
 
 print("success")
 

archive/lightning/train_predict.py

@@ -10,7 +10,7 @@ warnings.filterwarnings("ignore")
 import sys
 sys.path.insert(0, '..')
 
-from load_data import load_files
+from utils.load_data import load_files
 
 
 

archive/model_classification_keras.py

@@ -1,6 +1,6 @@
 #%% Import all the stuff, load data, define constants
 from sklearn.utils import shuffle
-from load_data import load_files, create_target_classes
+from utils.load_data import load_files, create_target_classes
 import pandas as pd
 from tensorflow import keras
 from utils.normalize import normalize

archive/model_classification_sklearn.py

@@ -2,7 +2,7 @@
 from sklearnex import patch_sklearn
 patch_sklearn()
 
-from load_data import create_target_cum_forward_returns, create_target_classes, load_files
+from utils.load_data import create_target_cum_forward_returns, create_target_classes, load_files
 from sktime.forecasting.model_selection import temporal_train_test_split
 from sklearn.metrics import accuracy_score
 from sklearn.metrics import confusion_matrix

archive/model_classification_sklearn_walk_forward.py

@@ -2,7 +2,7 @@
 from sklearnex import patch_sklearn
 patch_sklearn()
 
-from load_data import create_target_classes, load_files
+from utils.load_data import create_target_classes, load_files
 from sktime.forecasting.model_selection import temporal_train_test_split
 from sklearn.metrics import accuracy_score
 from sklearn.metrics import confusion_matrix
@@ -122,7 +122,7 @@ for model_name, create_model in models_to_try:
         retrain_every = retrain_every
     )
 
-    evaluate_predictions(model_name, y_train, preds, sliding_window_size)
+    evaluate_predictions(model_name, y_train, preds)
 
 
 #%%

archive/model_classification_sktime.py

@@ -1,5 +1,5 @@
 #%% Import all the stuff, load data, define constants
-from load_data import load_files, create_target_classes
+from utils.load_data import load_files, create_target_classes
 import pandas as pd
 import numpy as np
 from utils.sktime import from_df_to_sktime_data

archive/model_regression_keras.py

@@ -1,6 +1,6 @@
 #%% Import all the stuff, load data, define constants
 from sklearn.utils import shuffle
-from load_data import load_files, create_target_cum_forward_returns
+from utils.load_data import load_files, create_target_cum_forward_returns
 import pandas as pd
 from tensorflow import keras
 from utils.normalize import normalize

archive/run_lightning.py

@@ -1,4 +1,4 @@
-from load_data import load_files
+from utils.load_data import load_files
 import pandas as pd
 # from tensorflow import keras
 from utils.normalize import normalize

data/ADA_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.0243,0.02172,0.0223,0.02194
-2017-11-05,0.02283,0.02062,0.02194,0.02069
-2017-11-06,0.02324,0.01949,0.02069,0.02185
-2017-11-07,0.02379,0.02145,0.02185,0.02209
 2017-11-08,0.02754,0.02204,0.02209,0.02509
 2017-11-09,0.03493,0.02396,0.02509,0.03223
 2017-11-10,0.03014,0.02639,0.03223,0.02639
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,1.332,1.208,1.291,1.211
 2021-12-11,1.361,1.192,1.211,1.352
 2021-12-12,1.414,1.321,1.352,1.347
-2021-12-13,1.357,1.234,1.347,1.234
+2021-12-13,1.357,1.203,1.347,1.225
+2021-12-14,1.283,1.201,1.225,1.267
+2021-12-15,1.331,1.206,1.267,1.311
+2021-12-16,1.33,1.237,1.311,1.24
+2021-12-17,1.258,1.185,1.24,1.2

data/BNB_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,1.695,1.385,1.461,1.638
-2017-11-05,1.617,1.452,1.637,1.52
-2017-11-06,1.692,1.464,1.52,1.573
-2017-11-07,1.824,1.498,1.573,1.795
 2017-11-08,2.045,1.768,1.795,2.029
 2017-11-09,2.126,1.95,2.029,2.052
 2017-11-10,1.906,1.656,2.052,1.826
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,595.23,551.16,570.89,553.52
 2021-12-11,573.17,542.58,553.52,564.19
 2021-12-12,574.92,559.75,564.19,570.73
-2021-12-13,572.4,523.14,570.73,523.5
+2021-12-13,572.4,511.32,570.73,521.07
+2021-12-14,535.88,510.3,521.07,528.6
+2021-12-15,546.91,501.89,528.6,541.28
+2021-12-16,545.44,524.97,541.28,526.08
+2021-12-17,536.0,511.39,526.08,514.07

data/BTC_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,7492.24,6957.84,7146.98,7363.8
-2017-11-05,7599.44,7284.33,7363.8,7389.55
-2017-11-06,7429.7,6934.73,7389.55,6959.23
-2017-11-07,7222.37,6957.8,6959.27,7102.75
 2017-11-08,7869.1,7026.97,7102.23,7444.36
 2017-11-09,7457.12,7040.83,7444.36,7129.59
 2017-11-10,7318.97,6406.05,7129.59,6565.8
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,50047.85,46989.42,47593.85,47191.14
 2021-12-11,49504.07,46829.85,47191.14,49398.4
 2021-12-12,50807.75,48689.25,49398.4,50109.99
-2021-12-13,50215.42,47054.11,50109.99,47070.27
+2021-12-13,50215.42,45800.52,50109.99,46731.03
+2021-12-14,48665.63,46320.4,46731.03,48387.69
+2021-12-15,49499.91,46562.1,48387.69,48885.05
+2021-12-16,49451.4,47546.15,48885.05,47635.92
+2021-12-17,48004.37,45585.05,47635.92,45727.16

data/DOT_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.0,0.0,0.0,0.0
-2017-11-05,0.0,0.0,0.0,0.0
-2017-11-06,0.0,0.0,0.0,0.0
-2017-11-07,0.0,0.0,0.0,0.0
 2017-11-08,0.0,0.0,0.0,0.0
 2017-11-09,0.0,0.0,0.0,0.0
 2017-11-10,0.0,0.0,0.0,0.0
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,28.43,25.7,26.95,26.42
 2021-12-11,28.18,25.83,26.42,27.8
 2021-12-12,30.01,27.34,27.8,29.59
-2021-12-13,29.86,26.31,29.59,26.31
+2021-12-13,29.86,25.56,29.59,25.96
+2021-12-14,26.69,25.01,25.96,26.21
+2021-12-15,28.04,24.28,26.21,27.16
+2021-12-16,27.84,25.83,27.16,26.02
+2021-12-17,26.21,24.2,26.02,24.49

data/ETH_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,304.65,294.05,304.51,300.04
-2017-11-05,302.16,287.68,299.26,295.8
-2017-11-06,306.8,292.26,296.22,296.82
-2017-11-07,303.76,287.45,296.82,291.84
 2017-11-08,319.2,290.55,291.84,307.35
 2017-11-09,329.21,305.84,307.35,319.66
 2017-11-10,324.22,288.96,319.67,296.86
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,4229.29,3890.33,4111.92,3902.49
 2021-12-11,4099.04,3835.8,3902.49,4088.68
 2021-12-12,4179.55,3990.11,4088.68,4135.31
-2021-12-13,4144.68,3769.76,4135.31,3776.88
+2021-12-13,4144.68,3672.91,4135.31,3785.08
+2021-12-14,3882.64,3687.79,3785.08,3862.18
+2021-12-15,4095.28,3651.32,3862.18,4020.19
+2021-12-16,4113.2,3953.97,4020.19,3957.03
+2021-12-17,3994.8,3701.56,3957.03,3746.81

data/FIL_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.0,0.0,0.0,0.0
-2017-11-05,0.0,0.0,0.0,0.0
-2017-11-06,0.0,0.0,0.0,0.0
-2017-11-07,0.0,0.0,0.0,0.0
 2017-11-08,0.0,0.0,0.0,0.0
 2017-11-09,0.0,0.0,0.0,0.0
 2017-11-10,0.0,0.0,0.0,0.0
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,41.35,37.45,37.53,39.91
 2021-12-11,40.54,38.67,39.91,40.4
 2021-12-12,40.56,38.74,40.4,39.77
-2021-12-13,40.07,36.67,39.77,36.7
+2021-12-13,40.07,35.27,39.77,35.89
+2021-12-14,37.31,35.27,35.89,36.76
+2021-12-15,38.92,35.64,36.76,38.22
+2021-12-16,38.54,36.95,38.22,37.21
+2021-12-17,39.0,35.71,37.21,35.9

data/GLD.csv

@@ -1027,3 +1027,7 @@ time,open,high,low,close,volume
 2021-12-08,166.51,166.96,166.28,166.9,4288935.0
 2021-12-09,166.54,166.58,165.65,165.88,4913771.0
 2021-12-10,166.72,166.915,166.18,166.58,6450614.0
+2021-12-13,166.9,167.255,166.67,167.0,6610673.0
+2021-12-14,165.34,166.14,165.16,165.44,8637879.0
+2021-12-15,165.27,166.4,163.8,166.15,14231184.0
+2021-12-16,167.01,168.18,166.94,168.16,11186197.0

data/LTC_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,56.12,54.62,55.98,54.83
-2017-11-05,55.45,54.43,54.84,54.6
-2017-11-06,55.79,53.43,54.6,54.5
-2017-11-07,63.49,54.47,54.5,60.52
 2017-11-08,64.07,59.59,60.52,62.38
 2017-11-09,66.9,61.59,62.38,64.15
 2017-11-10,65.75,56.02,64.15,58.99
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,159.86,147.79,151.5,148.53
 2021-12-11,159.56,145.85,148.53,158.0
 2021-12-12,162.91,153.39,158.0,159.1
-2021-12-13,159.94,146.87,159.1,146.97
+2021-12-13,159.94,142.82,159.1,144.66
+2021-12-14,151.95,143.05,144.66,150.55
+2021-12-15,155.05,141.77,150.55,153.58
+2021-12-16,156.64,148.61,153.58,148.81
+2021-12-17,150.31,141.35,148.81,142.71

data/QQQ.csv

@@ -1027,3 +1027,7 @@ time,open,high,low,close,volume
 2021-12-08,398.17,399.82,396.22,399.61,39204041.0
 2021-12-09,398.35,400.2,393.5299,393.73,49973714.0
 2021-12-10,396.62,398.28,393.36,398.01,52545767.0
+2021-12-13,398.0,398.325,391.88,392.26,50166410.0
+2021-12-14,387.3,390.09,383.63,388.2,75199181.0
+2021-12-15,387.94,397.4198,383.88,397.05,73858028.0
+2021-12-16,398.04,398.475,385.21,386.84,82591368.0

data/SOL_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.0,0.0,0.0,0.0
-2017-11-05,0.0,0.0,0.0,0.0
-2017-11-06,0.0,0.0,0.0,0.0
-2017-11-07,0.0,0.0,0.0,0.0
 2017-11-08,0.0,0.0,0.0,0.0
 2017-11-09,0.0,0.0,0.0,0.0
 2017-11-10,0.0,0.0,0.0,0.0
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,184.56,167.31,181.41,167.65
 2021-12-11,173.8,161.48,167.65,172.08
 2021-12-12,176.87,167.8,172.08,173.0
-2021-12-13,173.0,155.92,173.0,156.29
+2021-12-13,173.0,148.1,173.0,155.15
+2021-12-14,163.17,149.67,155.15,161.43
+2021-12-15,183.17,156.91,161.43,178.55
+2021-12-16,189.3,174.51,178.55,176.92
+2021-12-17,182.47,168.15,176.92,169.95

data/SPY.csv

@@ -1027,3 +1027,7 @@ time,open,high,low,close,volume
 2021-12-08,468.7,470.0,466.83,469.52,72238777.0
 2021-12-09,468.15,469.6291,466.14,466.35,61272568.0
 2021-12-10,469.23,470.9,466.51,470.74,77159757.0
+2021-12-13,470.19,470.56,466.27,466.57,87724680.0
+2021-12-14,463.09,465.74,460.25,463.36,97264128.0
+2021-12-15,463.42,470.86,460.74,470.6,116899251.0
+2021-12-16,472.57,472.87,464.8,466.45,116568626.0

data/TRX_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.0026,0.00215,0.00217,0.00218
-2017-11-05,0.00255,0.00215,0.00218,0.00248
-2017-11-06,0.00248,0.00182,0.00248,0.00183
-2017-11-07,0.00234,0.00183,0.00183,0.00231
 2017-11-08,0.00247,0.00215,0.00231,0.00236
 2017-11-09,0.00246,0.00225,0.00236,0.00243
 2017-11-10,0.00243,0.002,0.00243,0.00228
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,0.09305,0.08754,0.08872,0.08778
 2021-12-11,0.0919,0.08668,0.08778,0.09154
 2021-12-12,0.09245,0.08986,0.09154,0.09109
-2021-12-13,0.09156,0.08604,0.09109,0.08606
+2021-12-13,0.09156,0.08423,0.09109,0.08504
+2021-12-14,0.08775,0.0839,0.08504,0.08716
+2021-12-15,0.08887,0.08298,0.08716,0.08758
+2021-12-16,0.08818,0.08529,0.08758,0.08542
+2021-12-17,0.08601,0.07859,0.08542,0.07927

data/UNI_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.0,0.0,0.0,0.0
-2017-11-05,0.0,0.0,0.0,0.0
-2017-11-06,0.0,0.0,0.0,0.0
-2017-11-07,0.0,0.0,0.0,0.0
 2017-11-08,0.0,0.0,0.0,0.0
 2017-11-09,0.0,0.0,0.0,0.0
 2017-11-10,0.0,0.0,0.0,0.0
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,16.53,15.12,16.2,15.16
 2021-12-11,16.74,14.97,15.16,15.88
 2021-12-12,16.2,15.28,15.88,16.03
-2021-12-13,16.09,14.44,16.03,14.47
+2021-12-13,16.09,13.77,16.03,14.16
+2021-12-14,15.08,13.93,14.16,15.0
+2021-12-15,15.51,14.17,15.0,15.2
+2021-12-16,15.37,14.36,15.2,14.42
+2021-12-17,14.58,13.68,14.42,13.86

data/XRP_USD.csv

@@ -1,8 +1,4 @@
 time,high,low,open,close
-2017-11-04,0.2061,0.2001,0.2059,0.2015
-2017-11-05,0.2033,0.1984,0.2014,0.1997
-2017-11-06,0.2067,0.1981,0.1997,0.2013
-2017-11-07,0.2065,0.1993,0.2013,0.2055
 2017-11-08,0.2235,0.2031,0.2055,0.2167
 2017-11-09,0.2218,0.2126,0.2167,0.2154
 2017-11-10,0.2185,0.1988,0.2154,0.2033
@@ -1499,4 +1495,8 @@ time,high,low,open,close
 2021-12-10,0.8823,0.7938,0.8594,0.7994
 2021-12-11,0.844,0.7857,0.7994,0.8387
 2021-12-12,0.8565,0.81,0.8387,0.8404
-2021-12-13,0.8444,0.7938,0.8404,0.7948
+2021-12-13,0.8444,0.7615,0.8404,0.7819
+2021-12-14,0.8215,0.7734,0.7819,0.8117
+2021-12-15,0.8398,0.7781,0.8117,0.8273
+2021-12-16,0.8364,0.8038,0.8273,0.8054
+2021-12-17,0.832,0.7766,0.8054,0.7831

fetch_data.py

@@ -1,26 +0,0 @@
-#%%
-import pandas as pd
-from utils.get_prices import get_crypto_price_crypto_compare, get_stock_price_av
-from itertools import combinations
-
-#%%
-crypto_tickers = ["BTC", "ETH", "BNB", "ADA", "SOL", "XRP", "DOT", "LTC", "UNI", "TRX", "FIL", "USD"]
-etf_tickers = ["GLD", "IEF", "TLT", "SPY", "QQQ"]
-tickers = crypto_tickers + etf_tickers
-
-#%%
-for ticker in etf_tickers:
-    print("Fetching ", ticker)
-    df = get_stock_price_av(ticker, "2017-11-10")
-
-    df.to_csv(f"data/{ticker}.csv", index=True)
-
-crypto_ticker_pairs = list(combinations(crypto_tickers, 2))
-for src_ticker, trg_ticker in crypto_ticker_pairs:
-    print("Fetching ", src_ticker, trg_ticker)
-    df = get_crypto_price_crypto_compare(src_ticker, trg_ticker, 1500)
-
-    df.to_csv(f"data/{src_ticker}_{trg_ticker}.csv", index=True)
-
-
-# %%

run_fetch_data.py

@@ -0,0 +1,22 @@
+#%%
+import pandas as pd
+from utils.get_prices import get_crypto_price_crypto_compare, get_stock_price_av
+
+#%%
+crypto_tickers = ["BTC", "ETH", "BNB", "ADA", "SOL", "XRP", "DOT", "LTC", "UNI", "TRX", "XLM", "BCH", "FIL", "ETC", "THETA", "XTZ"]
+etf_tickers = ["GLD", "IEF", "TLT", "SPY", "QQQ"]
+
+#%%
+for ticker in etf_tickers:
+    print("Fetching ", ticker)
+    df = get_stock_price_av(ticker, "2017-11-10")
+
+    df.to_csv(f"data/{ticker}.csv", index=True)
+
+for src_ticker in crypto_tickers:
+    print("Fetching ", src_ticker, "USD")
+    df = get_crypto_price_crypto_compare(src_ticker, "USD", 1500)
+
+    df.to_csv(f"data/{src_ticker}_USD.csv", index=True)
+
+

run_pipeline.py

@@ -1,7 +1,7 @@
 from sklearnex import patch_sklearn
 patch_sklearn()
 
-from load_data import get_crypto_assets, get_etf_assets, load_data
+from utils.load_data import get_crypto_assets, get_etf_assets, load_data
 
 import pandas as pd
 import numpy as np
@@ -19,43 +19,44 @@ from training.pipeline import run_single_asset_trainig_pipeline
 
 # Parameters
 regression_models = [
-    ('Lasso', Lasso(alpha=1.0, max_iter=10000)),
-    ('Ridge', Ridge(alpha=1.0)),
+    # ('Lasso', Lasso(alpha=0.1, max_iter=1000)),
+    ('Ridge', Ridge(alpha=0.1)),
     ('BayesianRidge', BayesianRidge()),
-    ('KNN', KNeighborsRegressor(n_neighbors=15)),
-    # ('AB', AdaBoostRegressor()),
+    ('KNN', KNeighborsRegressor(n_neighbors=25)),
+    # ('AB', AdaBoostRegressor(random_state=1)),
     # ('LR', LinearRegression(n_jobs=-1)),
     # ('MLP', MLPRegressor(hidden_layer_sizes=(100,20), max_iter=1000)),
     # ('RF', RandomForestRegressor(n_jobs=-1)),
     # ('SVR', SVR(kernel='rbf', C=1e3, gamma=0.1))
 ]
-ensemble_model = [('Ensemble - Lasso', Lasso(alpha=1.0, max_iter=10000, positive=True))]
+regression_ensemble_model = [('Ensemble - Ridge', Ridge(alpha=0.1))]
 
 classification_models = [
     ('LR', LogisticRegression(n_jobs=-1)),
-    # ('LDA', LinearDiscriminantAnalysis()),
-    # ('KNN', KNeighborsClassifier()),
-    # ('CART', DecisionTreeClassifier()),
-    # ('NB', GaussianNB()),
+    ('LDA', LinearDiscriminantAnalysis()),
+    ('KNN', KNeighborsClassifier()),
+    ('CART', DecisionTreeClassifier()),
+    ('NB', GaussianNB()),
     # ('AB', AdaBoostClassifier()),
     # ('RF', RandomForestClassifier(n_jobs=-1))
 ]
+classification_ensemble_model = [('Ensemble - CART', DecisionTreeClassifier())]
 
 
 path = 'data/'
 all_assets = get_crypto_assets(path)
 
-sliding_window_size = 200
-retrain_every = 100
-scaler = 'none' # 'normalize' 'minmax' 'standardize' 'none'
+sliding_window_size = 150
+retrain_every = 20
+scaler = 'minmax' # 'normalize' 'minmax' 'standardize' 'none'
 include_original_data_in_ensemble = True
-method = 'regression'
+method = 'classification'
 data_parameters = dict(path=path,
     target_asset_lags= [1,2,3,4,5,6,8,10,15],
-    load_other_assets= True,
+    load_other_assets= False,
     other_asset_lags= [],
     log_returns= True,
-    add_date_features= True,
+    add_date_features= False,
     own_technical_features= 'level2',
     other_technical_features= 'none',
     exogenous_features= 'none',
@@ -101,10 +102,10 @@ for asset in all_assets:
     ensemble_result, ensemble_preds = run_single_asset_trainig_pipeline(
         ticker_to_predict = asset,
         X = ensemble_X,
-        y = target_returns,
+        y = y,
         target_returns = target_returns,
-        models = ensemble_model,
-        method = 'regression',
+        models = regression_ensemble_model if method == 'regression' else classification_ensemble_model,
+        method = method,
         sliding_window_size = sliding_window_size,
         retrain_every = retrain_every,
         scaler = scaler

tests/test_walk_forward.py

@@ -1,4 +1,3 @@
-import pytest
 import numpy as np
 import pandas as pd
 from training.walk_forward import walk_forward_train_test

training/pipeline.py

@@ -50,7 +50,6 @@ def run_single_asset_trainig_pipeline(
             model_name = model_name,
             target_returns = target_returns,
             y_pred = preds,
-            sliding_window_size = sliding_window_size,
             method = method,
         )
         column_name = ticker_to_predict + "_" + model_name

utils/evaluate.py

@@ -5,7 +5,7 @@ from utils.helpers import get_first_valid_return_index
 import pandas as pd
 import numpy as np
 
-def backtest(returns: pd.Series, signal: pd.Series, transaction_cost = 0.01) -> pd.Series:
+def backtest(returns: pd.Series, signal: pd.Series, transaction_cost = 0.00) -> pd.Series:
     delta_pos = signal.diff(1).abs().fillna(0.)
     costs = transaction_cost * delta_pos
     return (signal * returns) - costs
@@ -36,20 +36,18 @@ def evaluate_predictions(
                         model_name: str,
                         target_returns: pd.Series,
                         y_pred: pd.Series,
-                        sliding_window_size: int,
                         method: Literal['classification', 'regression']
                         ) -> pd.Series:
     # ignore the predictions until we see a non-zero returns (and definitely skip the first sliding_window_size)
-    first_nonzero_return = get_first_valid_return_index(target_returns)
-    evaluate_from = first_nonzero_return + sliding_window_size + 1
+    first_nonzero_return = max(get_first_valid_return_index(target_returns), get_first_valid_return_index(y_pred))
+    evaluate_from = first_nonzero_return + 1
     
     target_returns = pd.Series(target_returns[evaluate_from:])
     if method == 'regression':
         # if there are lots of zeros in the ground truth returns, probably something is wrong, but we can tolerate a couple of days of missing data.
         is_zero = target_returns[target_returns == 0]
         assert len(is_zero) < 15
-    # we can't deal with 0 returns, so we'll just remap the few examples to 0.0001 
-    target_returns = target_returns.apply(lambda x: 0.0001 if x == 0 else x)
+
     y_pred = pd.Series(y_pred[evaluate_from:])
 
     df = __preprocess(target_returns, y_pred, method)

utils/helpers.py

@@ -1,6 +1,6 @@
+from typing import Literal
 import pandas as pd
 import numpy as np
 
 def get_first_valid_return_index(series: pd.Series) -> int:
-    first_nonzero_return = np.where(np.logical_and(series != 0, np.logical_not(np.isnan(series))))[0][0]
-    return first_nonzero_return
+    return np.where(np.logical_and(series != 0, np.logical_not(np.isnan(series))))[0][0]

utils/load_data.py

@@ -187,7 +187,7 @@ def __create_target_classes(df: pd.DataFrame, source_column: str, period: int, n
     assert period > 0
 
     def get_class_binary(x):
-        return 0 if x <= 0.0 else 1
+        return -1 if x <= 0.0 else 1
 
     def get_class_threeway(x):
         bins = pd.qcut(df[source_column], 4, duplicates='raise', retbins=True)[1]