我利用上述代码进行顺序向前选择时，显示出的评价指标都很完美

Question

# 分离特征和目标变量
y = '23h结果'
X= [x for x in data.columns if x not in [y, '23h结果']]
# 初始化测试集
X_test = data[X].sample(frac=0.2, random_state=42)
y_test = data[y].loc[X_test.index]

# 初始化评价指标列表
mae_list = []
mse_list = []
rmse_list = []
r2_list = []

# 初始化选中特征列表
selected_features = []

for i in range(len(X)):
    mae_min = float('inf')
    mse_min = float('inf')
    rmse_min = float('inf')
    r2_max = float('-inf')
    best_feature = None

    for feature in X:
        # 若特征已经被选中则跳过
        if feature in selected_features:
            continue

        # 加入待选特征
        cur_features = selected_features + [feature]

        # 划分训练集和验证集
        X_train = data[cur_features].dropna()
        y_train = data[y].loc[X_train.index]

        # 构建模型
        model = lgb.LGBMRegressor(random_state=42)
        model.fit(X_train, y_train)

        # 评估模型
        y_pred = model.predict(X_test[cur_features])
        mae = mean_absolute_error(y_test, y_pred)
        mse = mean_squared_error(y_test, y_pred)
        rmse = mean_squared_error(y_test, y_pred, squared=False)
        r2 = r2_score(y_test, y_pred)

        # 记录最佳特征
        if mae < mae_min:
            mae_min = mae
            best_feature = feature
        if mse < mse_min:
            mse_min = mse
        if rmse < rmse_min:
            rmse_min = rmse
        if r2 > r2_max:
            r2_max = r2

    # 记录评价指标
    mae_list.append(mae_min)
    mse_list.append(mse_min)
    rmse_list.append(rmse_min)
    r2_list.append(r2_max)

    # 记录选中特征
    selected_features.append(best_feature)

    # 输出评价指标
    print(f'Round {i + 1}:')
    print(f'Selected Feature: {best_feature}')
    print(f'MAE: {mae_min:.4f}')
    print(f'MSE: {mse_min:.4f}')
    print(f'RMSE: {rmse_min:.4f}')
    print(f'R2: {r2_max:.4f}\n')

我利用上述代码进行顺序向前选择时，为什么我无论用什么数据，显示出的评价指标都很完美，是哪里出现问题了呢？

Answer 1

可能数据集较小或者本身挺好