generate label for mysql

main_spider.py

@@ -72,41 +72,39 @@ class LightGBM(object):
             features[key] = self.label_encoder.fit_transform(features[key])
         return features, labels
 
-    def bays(self):
-        # 创建LightGBM数据集，注意不要在这里指定categorical_feature，因为我们使用的是玩具数据集
+    def objective(self, trial):
         x, y = self.read_data()
-        train_size = int(len(x) * 0.9)
+        train_size = int(len(x) * self.split_c)
         X_train, X_test = x[:train_size], x[train_size:]
         Y_train, Y_test = y[:train_size], y[train_size:]
-        train_data = lgb.Dataset(X_train, label=Y_train)
-        def lgbm_eval(num_leaves, learning_rate, feature_fraction, bagging_fraction, bagging_freq, min_child_samples):
-            params = {
-                'objective': 'binary',
-                'metric': 'auc',
-                'verbose': -1,
-                'num_leaves': int(num_leaves),
-                'learning_rate': learning_rate,
-                'feature_fraction': feature_fraction,
-                'bagging_fraction': bagging_fraction,
-                'bagging_freq': int(bagging_freq),
-                'min_child_samples': int(min_child_samples),
-            }
-            cv_result = lgb.cv(params, train_data, nfold=5, seed=42, stratified=True, metrics=['auc'])
-            return max(cv_result['auc-mean'])
-
-        param_bounds = {
-            'num_leaves': (20, 40),
-            'learning_rate': (1e-4, 1e-2),
-            'feature_fraction': (0.5, 0.8),
-            'bagging_fraction': (0.5, 0.8),
-            'bagging_freq': (1, 10),
-            'min_child_samples': (20, 100),
+
+        dtrain = lgb.Dataset(X_train, label=Y_train)
+        dvalid = lgb.Dataset(X_test, label=Y_test, reference=dtrain)
+
+        param = {
+            'objective': 'binary',  # 根据问题修改，例如'regression'或'multiclass'
+            'metric': 'binary_logloss',  # 根据问题修改，例如'l2'或'multi_logloss'
+            'verbosity': -1,
+            'boosting_type': 'gbdt',
+            'num_leaves': trial.suggest_int('num_leaves', 20, 40),
+            'learning_rate': trial.suggest_float('learning_rate', 1e-4, 1e-1, log=True),
+            'feature_fraction': trial.suggest_float('feature_fraction', 0.6, 0.9),
+            'bagging_fraction': trial.suggest_float('bagging_fraction', 0.6, 0.9),
+            'bagging_freq': trial.suggest_int('bagging_freq', 1, 10),
+            'min_child_samples': trial.suggest_int('min_child_samples', 5, 100),
         }
 
-        optimizer = BayesianOptimization(f=lgbm_eval, pbounds=param_bounds, random_state=42)
-        optimizer.maximize(init_points=5, n_iter=25)
+        gbm = lgb.train(param, dtrain, valid_sets=[dvalid], early_stopping_rounds=100, verbose_eval=False)
+        preds = gbm.predict(X_test)
+        pred_labels = np.rint(preds)
+        accuracy = accuracy_score(Y_test, pred_labels)
+        return accuracy  # 或其他优化指标
 
-        print("Best Parameters:", optimizer.max['params'])
+    def tune(self, n_trials=100):
+        study = optuna.create_study(direction='maximize')
+        study.optimize(self.objective, n_trials=n_trials)
+        print('Number of finished trials:', len(study.trials))
+        print('Best trial:', study.best_trial.params)
 
     def train_model(self):
         """
@@ -212,7 +210,7 @@ if __name__ == "__main__":
     #     L.evaluate_model()
     #     L.feature_importance()
     L = LightGBM("train", "whole")
-    L.bays()
+    L.tune()
     # study = optuna.create_study(direction='maximize')
     # study.optimize(L.bays_params, n_trials=100)
     # print('Number of finished trials:', len(study.trials))