From 495857913d0a1c727fa8aab51e11fbe43110251b Mon Sep 17 00:00:00 2001
From: duxin <duxin@example.com>
Date: Wed, 29 Jul 2026 10:28:17 +0800
Subject: [PATCH] =?UTF-8?q?refactor:=20RandomizedSearchCV=20=E2=86=92=20Gr?=
 =?UTF-8?q?idSearchCV=20=E2=80=94=20SVR=20=E8=B6=85=E5=8F=82=E5=85=A8?=
 =?UTF-8?q?=E9=87=8F=E7=A9=B7=E4=B8=BE?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- SVR 参数网格仅 216 种组合（4×6×3×3），全量搜索 ~2s 完成
- 之前 n_iter=10 只抽样 4.6%，靠运气撞最优参数
- 删除 n_iter 和 random_state 参数（GridSearchCV 不需要）
- 保留 RandomizedSearchCV import 供其他模型（参数空间大的）使用
---
 src/core/modeling/modeling_batch.py | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/src/core/modeling/modeling_batch.py b/src/core/modeling/modeling_batch.py
index 5e27447..83f4eda 100644
--- a/src/core/modeling/modeling_batch.py
+++ b/src/core/modeling/modeling_batch.py
@@ -14,7 +14,7 @@ from sklearn.svm import SVR
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.neighbors import KNeighborsRegressor
 from sklearn.linear_model import LinearRegression, Ridge, Lasso, ElasticNet
-from sklearn.model_selection import RandomizedSearchCV, cross_val_score, KFold, train_test_split
+from sklearn.model_selection import GridSearchCV, RandomizedSearchCV, cross_val_score, KFold, train_test_split
 from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
 from sklearn.cross_decomposition import PLSRegression
 from sklearn.ensemble import GradientBoostingRegressor, AdaBoostRegressor, ExtraTreesRegressor
@@ -646,23 +646,22 @@ class WaterQualityModelingBatch:
         X_train[~np.isfinite(X_train)] = np.nan
         X_test[~np.isfinite(X_test)]   = np.nan
 
-        # RandomizedSearchCV 需要以「步骤名__参数名」的格式索引参数网格；
-        # 我们原有的 config['params'] 是模型层的（无 __），统一加 model__ 前缀。
+        # 以「步骤名__参数名」的格式索引参数网格；
+        # config['params'] 是模型层的（无 __），统一加 model__ 前缀。
         prefixed_params = {
             f"model__{k}": v for k, v in config['params'].items()
         }
 
-        # 随机搜索：直接对 Pipeline 调优（scaler 仅在 train fold 上 fit）
+        # 全量网格搜索：SVR 超参组合仅 216 种（4×6×3×3），
+        # 穷举远优于 RandomizedSearchCV(n_iter=10) 的随机抽样
         cv_strategy = KFold(n_splits=cv_folds, shuffle=True, random_state=random_state)
 
-        grid_search = RandomizedSearchCV(
+        grid_search = GridSearchCV(
             pipeline,
             prefixed_params,
-            n_iter=10,
             cv=cv_strategy,
             scoring=scoring,
             n_jobs=safe_n_jobs,
-            random_state=random_state,
             verbose=1,
         )