网格搜索后改善不大

# Instantiate random forest
from sklearn.ensemble import RandomForestClassifier
random_forest = RandomForestClassifier()

# K-Fold Cross Validation (for grid search)
inner_cross_validator = StratifiedKFold(n_splits=k_fold, shuffle=True)


from sklearn.model_selection import GridSearchCV

# Define parameters for grid search
number_of_trees = {'n_estimators': [100, 300, 500]}
max_features_per_tree = {'max_features': [0.2, 0.5, 0.8]}
min_samples_split_per_node = {'min_samples_split': [0.2, 0.5, 0.8]}
parameters = {**number_of_trees, **max_features_per_tree, **min_samples_split_per_node}

# Execute grid search and retrieve the best classifier
best_random_forest = GridSearchCV(estimator=random_forest, param_grid=parameters, scoring='average_precision', cv=inner_cross_validator, n_jobs=3)
best_random_forest.fit(X_train, y_train)

然而,网格搜索后的查全率和查全率几乎没有提高。

一般来说,在我过去处理其他数据集的经验中,我没有注意到在网格搜索之后,与SkLearn这样的库的默认值相比,在各种度量的得分上有超过5%或很少有10%的改进。

我能做些什么(在特征工程阶段之后)来显著提高我的分类模型的性能吗?