将前向特征选择与 KNN 结合时出现 InvalidIndexError: (slice(None, None, None), [0])

Question

我正在尝试使用 uci 心脏病数据集来预测心脏病。我正在应用前向特征选择技术来选择特征，然后应用 knn 进行预测。附上代码片段。

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
selected_features = set()

# Initialize a KNN classifier
knn_classifier = KNeighborsClassifier(n_neighbors=5)

# Number of features to select (you can choose based on your requirements)
num_features_to_select = 10

# Perform forward feature selection
for _ in range(num_features_to_select):
    best_accuracy = 0.0
    best_feature = None
    
    # Iterate over remaining features
    for feature_index in range(X_train.shape[1]):
        if feature_index not in selected_features:
            # Add the feature and train the KNN classifier
            current_features = list(selected_features) + [feature_index]
            knn_classifier.fit(X_train[:, current_features], y_train)
            
            # Predict using the trained KNN classifier
            y_pred = knn_classifier.predict(X_test[:, current_features])
            
            # Calculate accuracy
            accuracy = accuracy_score(y_test, y_pred)
            
            # Update the best feature and accuracy
            if accuracy > best_accuracy:
                best_accuracy = accuracy
                best_feature = feature_index
    
    # Add the best feature to the selected features
    selected_features.add(best_feature)
    
    # Print the selected feature and its accuracy
    print(f"Selected Feature: {best_feature}, Accuracy: {best_accuracy}")

# Train the KNN classifier with the selected features
knn_classifier.fit(X_train[:, list(selected_features)], y_train)

# Predict using the trained KNN classifier with the selected features
y_pred = knn_classifier.predict(X_test[:, list(selected_features)])

# Calculate accuracy
accuracy = accuracy_score(y_test, y_pred)
print("Final Accuracy:", accuracy)

但是我收到错误为

InvalidIndexError: (slice(None, None, None), [0])

我做错了什么？

Answer 1

 # Iterate over remaining features
    for feature_index in range(X_train.shape[1]):
        if feature_index not in selected_features:
            # Add the feature and train the KNN classifier
            current_features = list(selected_features) + [feature_index]
            knn_classifier.fit(X_train[:, current_features], y_train)

在此 for 循环中，feature_index 获取的第一个值是 0。您不是迭代列名，而是迭代一个范围。所以，当你检查

if feature_index not in selected_features

时，它没有意义。请检查这部分。