adding predictors.pdf

models/model_utils.py

@@ -57,11 +57,9 @@ def scaling_handler(data_frame, method="robust_scaling"):
         StandardScaler,
     )
 
-    # Separate features and label
     labels = data_frame["label"]
     X = data_frame.drop("label", axis=1)
 
-    # Choose scaler/transformer
     if method == "robust_scaling":
         scaler = RobustScaler()
     elif method == "standard_scaling":
@@ -87,13 +85,12 @@ def scaling_handler(data_frame, method="robust_scaling"):
 
         for col in X_pos.columns:
             if min_per_column[col] <= 0:
-                X_pos[col] = X_pos[col] + abs(min_per_column[col]) + 1e-6  # tiny offset
+                X_pos[col] = X_pos[col] + abs(min_per_column[col]) + 1e-6
 
         X = X_pos
     else:
         raise ValueError(f"Unknown scaling method: {method}")
 
-    # Fit and transform
     X_scaled = scaler.fit_transform(X)
     data_frame_scaled = pandas.DataFrame(X_scaled, columns=X.columns)
     data_frame_scaled["label"] = labels.values
@@ -101,16 +98,15 @@ def scaling_handler(data_frame, method="robust_scaling"):
     return data_frame_scaled
 
 
-from sklearn.metrics import (
-    accuracy_score,
-    f1_score,
-    fbeta_score,
-    precision_score,
-    recall_score,
-)
-
-
 def get_metrics(y_true, y_pred, prefix=""):
+    from sklearn.metrics import (
+        accuracy_score,
+        f1_score,
+        fbeta_score,
+        precision_score,
+        recall_score,
+    )
+
     metrics = {}
     metrics[f"{prefix}accuracy"] = accuracy_score(y_true, y_pred)
     metrics[f"{prefix}f1_macro"] = f1_score(y_true, y_pred, average="macro")
@@ -120,7 +116,6 @@ def get_metrics(y_true, y_pred, prefix=""):
         y_true, y_pred, average="macro"
     )
 
-    # Per-class scores
     f1_scores = f1_score(y_true, y_pred, average=None, zero_division=0)
     f2_scores = fbeta_score(y_true, y_pred, beta=2, average=None, zero_division=0)
     recall_scores = recall_score(y_true, y_pred, average=None, zero_division=0)
@@ -132,7 +127,6 @@ def get_metrics(y_true, y_pred, prefix=""):
         metrics[f"{prefix}recall_class{i}"] = recall_scores[i]
         metrics[f"{prefix}precision_class{i}"] = precision_scores[i]
 
-    # Confusion-matrix components
     TP = sum((y_true == 1) & (y_pred == 1))
     TN = sum((y_true == 0) & (y_pred == 0))
     FP = sum((y_true == 0) & (y_pred == 1))
@@ -151,24 +145,13 @@ import pandas as pd
 
 
 def average_fold_results(fold_results):
-    """
-    Computes the average of metrics over multiple folds.
-
-    fold_results: list of dictionaries, each containing metrics for one fold
-
-    Returns:
-        dict of average metrics
-    """
     if not fold_results:
         return {}
 
-    # Convert list of dicts to DataFrame
     df = pd.DataFrame(fold_results)
 
-    # Compute mean for each column
     avg_metrics = df.mean().to_dict()
 
-    # Convert any NumPy types to float
     for k, v in avg_metrics.items():
         if isinstance(v, (np.float32, np.float64)):
             avg_metrics[k] = float(v)