使用BorutaPy进行特征选择并输出带标题的Pandas DataFrame

import numpy as np
import pandas as pd
from sklearn.ensemble import RandomForestRegressor
from boruta import BorutaPy

# 假设 X 是特征DataFrame，y 是目标变量Series
# 为了演示，我们先创建一些示例数据
np.random.seed(42)
data_size = 100
X = pd.DataFrame(np.random.rand(data_size, 10), columns=[f'Feature_{i}' for i in range(10)])
# 假设 'Feature_0', 'Feature_2', 'Feature_5' 是相关特征
y = X['Feature_0'] * 2 + X['Feature_2'] * 3 - X['Feature_5'] * 0.5 + np.random.randn(data_size) * 0.1

# 初始化基线模型
model = RandomForestRegressor(n_estimators=100, max_depth=5, random_state=42)

# 初始化Boruta特征选择器
feat_selector = BorutaPy(
    verbose=2, # 详细输出模式
    estimator=model,
    n_estimators='auto',
    max_iter=50, # 迭代次数
    random_state=42,
)

# 训练Boruta。注意：BorutaPy的fit方法通常期望numpy数组作为输入
feat_selector.fit(X.values, y.values)

# 打印每个特征的支持度和排名
print("\n------Support and Ranking for each feature------\n")
for i in range(len(feat_selector.support_)):
    if feat_selector.support_[i]:
        print(f"Passes the test: {X.columns[i]} - Ranking: {feat_selector.ranking_[i]} ✔️")
    else:
        print(f"Doesn't pass the test: {X.columns[i]} - Ranking: {feat_selector.ranking_[i]} ❌")

# 获取筛选后的特征数组
X_filtered_array = feat_selector.transform(X.values)

# 步骤1：获取被选中的特征列名
selected_feature_names = X.columns[feat_selector.support_]

# 步骤2：使用transform方法获取筛选后的特征数据（NumPy数组）
# 注意：transform方法也期望NumPy数组作为输入
X_filtered_data = feat_selector.transform(X.values)

# 步骤3：将数据和列名组合成一个新的Pandas DataFrame
X_filtered_df_method1 = pd.DataFrame(X_filtered_data, columns=selected_feature_names)

print("\n--- Method 1: Manually Constructed DataFrame ---")
print(X_filtered_df_method1.head())
print(f"Shape of X_filtered_df_method1: {X_filtered_df_method1.shape}")

# 注意：此方法要求BorutaPy库版本支持 `return_df=True` 参数。
# 建议通过 `pip install boruta-py --upgrade` 更新到最新版本。

try:
    # 尝试使用 return_df=True 参数
    X_filtered_df_method2 = feat_selector.transform(X, return_df=True)
    print("\n--- Method 2: Using return_df=True ---")
    print(X_filtered_df_method2.head())
    print(f"Shape of X_filtered_df_method2: {X_filtered_df_method2.shape}")
except TypeError as e:
    print(f"\n--- Method 2: Not available ---")
    print(f"Error: {e}. Your BorutaPy version might not support `return_df=True`.")
    print("Please consider updating BorutaPy (`pip install boruta-py --upgrade`) or use Method 1.")