feat(Core): ensemble models, correct forward returns calculation, scaling, only train from when asset returns are available, major bug fixed in walk_forward_train_test (#35)

* fix(Core): correct forward returns calculation, classifiers are now working again, only train from when asset returns are available * feat(Utils): added get_first_valid_return_index() * feat(Ensemble): return models from `run_whole_pipeline` * feat(Ensemble): added ensemble step, fixed walk_forward_train_test predictions index confusion, * chore(Pipeline): remove unnecessary extra ensemble results dataframe * refactor(Core): removed unnecessary ensemble_train_predict, moved run_single_asset_trainig_pipeline to a separate file * feat(Training): added scaling on expanding window (the past) to walk_forward_train_test(), now printing out mean sharpe ratio * feat(CI): added environment.yml file * chore(Environment): update env.yml * feat(CI): added testing workflow * fix(CI): renamed enviroment.yml * fix(Tests): added missing new parameter to walk_forward_train_test()
2026-08-17 21:08:08 +00:00 · 2021-12-17 14:32:17 +01:00
parent 1eaba0c221
commit cc7061b456
12 changed files with 385 additions and 195 deletions
@@ -0,0 +1,63 @@
+import pandas as pd
+from sklearn.base import clone
+from utils.typing import SKLearnModel
+import numpy as np
+from utils.helpers import get_first_valid_return_index
+
+def walk_forward_train_test(
+                            model_name: str,
+                            model: SKLearnModel,
+                            X: pd.DataFrame,
+                            y: pd.Series,
+                            target_returns: pd.Series,
+                            window_size: int,
+                            retrain_every: int,
+                            scaler,
+                        ) -> tuple[pd.Series, pd.Series]:
+    assert len(X) == len(y)
+    predictions = pd.Series(index=y.index).rename(model_name)
+    models = pd.Series(index=y.index).rename(model_name)
+
+    first_nonzero_return = max(get_first_valid_return_index(target_returns), get_first_valid_return_index(X.iloc[:,0]))
+    train_from = first_nonzero_return + window_size + 1
+    train_till = len(y)
+    
+    iterations_since_retrain = 0
+    if scaler is not None:
+        scaler = clone(scaler)
+
+    for index in range(train_from, train_till):
+
+        iterations_since_retrain += 1
+
+        if iterations_since_retrain >= retrain_every or pd.isna(models[index-1]):
+            train_window_start = index - window_size - 1
+            train_window_end = index - 1
+            
+            if scaler is not None:
+                # First we need to fit on the expanding window data slice
+                # This is our only way to avoid lookahead bia
+                X_expanding_window = X[first_nonzero_return:train_window_end]
+                scaler.fit(X_expanding_window)
+
+            X_slice = X[train_window_start:train_window_end]
+            y_slice = y[train_window_start:train_window_end]
+
+            if scaler is not None:
+                X_slice = scaler.transform(X_slice)
+            else:
+                X_slice = X_slice.to_numpy()
+
+            current_model = clone(model)
+            current_model.fit(X_slice, y_slice.to_numpy())
+            iterations_since_retrain = 0
+        else:
+            current_model = models[index-1]
+
+        models[index] = current_model
+
+        next_timestep = X.iloc[index].to_numpy().reshape(1, -1)
+        prediction = current_model.predict(next_timestep).item()
+        predictions[index] = prediction
+
+    return models, predictions