如何在 scikit-learn 中将 SelectFromModel 与 cross_validate 结合使用？

Question

在训练 DecisionTreeClassifier 模型之前，我尝试使用 SelectFromModel 从数据集中选择特征。我还使用 cross_validate 来评估模型性能。但是，我不知道如何一起使用这两个功能。

import numpy as np
from sklearn.feature_selection import SelectFromModel
from sklearn.ensemble import RandomForestClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import cross_validate


# Split the dataset into features and target
X = dataset.drop('target_column', axis=1)  # Remove 'target_column' with the actual target column name
y = dataset['target_column']

# Feature selection using SelectFromModel(RandomForestClassifier)
selector = SelectFromModel(RandomForestClassifier(n_estimators=100, random_state=42))
X = selector.fit_transform(X, y)
num_selected_features = X.shape[1]
print(f'Number of Features Selected: {num_selected_features}')

# train a DecisionTreeClassifier with specificity as the evaluation metric
classifier = DecisionTreeClassifier(criterion="entropy",
                                   min_samples_split=5,
                                   random_state=0)

# Cross-validate the model using specificity as the evaluation metric
cross_validation_results = cross_validate(estimator=classifier,
                                        X=X,
                                        y=y,
                                        cv=5,
                                        scoring="accuracy",
                                        return_train_score=True)

# Print the cross-validation results
print(cross_validation_results)

Answer 1

from sklearn.pipeline import Pipeline
from sklearn.model_selection import cross_validate

# Define your feature selector and classifier
selector = SelectFromModel(RandomForestClassifier(n_estimators=100, random_state=42))
classifier = DecisionTreeClassifier(criterion="entropy", min_samples_split=5, random_state=0)

# Create a pipeline that first applies the feature selection and then runs the classifier
pipeline = Pipeline([
    ('feature_selection', selector),
    ('classification', classifier)
])

# Now, run cross-validation on the pipeline
cross_validation_results = cross_validate(estimator=pipeline,
                                          X=X,
                                          y=y,
                                          cv=5,
                                          scoring="accuracy",
                                          return_train_score=True)

# Print the cross-validation results
print(cross_validation_results)

这样，特征选择将应用于交叉验证中的每个训练折叠，避免潜在的数据泄漏。