feat: implement optimal threshold calculation for binary predictions in train function

app/training/train_classifier.py

@@ -170,7 +170,8 @@ def train(
                 cv=cv,
                 method="predict_proba",
             )[:, 1]
-        y_pred = (y_prob >= 0.5).astype(int)
+        threshold = _optimal_threshold(y, y_prob)
+        y_pred = (y_prob >= threshold).astype(int)
         cv_time = time.time() - t0
 
         acc = accuracy_score(y, y_pred)
@@ -194,6 +195,7 @@ def train(
             "recall": round(rec, 4),
             "f1": round(f1, 4),
             "roc_auc": round(auc, 4),
+            "optimal_threshold": round(threshold, 4),
             "validation_auc": round(tuning_meta.get("validation_auc", 0.0), 4),
             "selection_time_sec": round(tuning_meta.get("selection_time_sec", 0.0), 2),
             "train_time_sec": round(cv_time, 2),