feat: update default LightGBM params to Optuna best (trial #46, AUC=0.6002)

Optuna 50 trials Walk-Forward 5폴드 탐색 결과 (tune_results_20260302_144749.json):
- Baseline AUC: 0.5803 → Best AUC: 0.6002 (+0.0199, +3.4%)
- n_estimators: 500 → 434
- learning_rate: 0.05 → 0.123659
- max_depth: (미설정) → 6
- num_leaves: 31 → 14
- min_child_samples: 15 → 10
- subsample: 0.8 → 0.929062
- colsample_bytree: 0.8 → 0.946330
- reg_alpha: 0.05 → 0.573971
- reg_lambda: 0.1 → 0.000157
- weight_scale: 1.0 → 1.783105

Made-with: Cursor

scripts/train_model.py

@@ -151,16 +151,17 @@ def _load_lgbm_params(tuned_params_path: str | None) -> tuple[dict, float]:
     반환: (lgbm_params, weight_scale)
     """
     lgbm_params: dict = {
-        "n_estimators":      500,
-        "learning_rate":     0.05,
-        "num_leaves":        31,
-        "min_child_samples": 15,
-        "subsample":         0.8,
-        "colsample_bytree":  0.8,
-        "reg_alpha":         0.05,
-        "reg_lambda":        0.1,
+        "n_estimators":      434,
+        "learning_rate":     0.123659,
+        "max_depth":         6,
+        "num_leaves":        14,
+        "min_child_samples": 10,
+        "subsample":         0.929062,
+        "colsample_bytree":  0.946330,
+        "reg_alpha":         0.573971,
+        "reg_lambda":        0.000157,
     }
-    weight_scale = 1.0
+    weight_scale = 1.783105
 
     if tuned_params_path:
         with open(tuned_params_path, "r", encoding="utf-8") as f: