再利用pearson系数，找到对光伏发电功率影响最大的(3-4)个特征，并重新完成上面的3-5项

Pearson系数是用来衡量两个变量之间的线性相关程度，适合于分析数值型特征间的关联。为了找出对光伏发电功率影响最大的3到4个特征，我们可以按如下步骤操作：

1. 计算特征与发电功率的Pearson系数：

from scipy.stats import pearsonr
import pandas as pd

def calculate_pearson_coefficients(data, target_column):
    correlations = {}
    for column in data.columns[:-1]:  # 不包括最后一列（目标变量）
        corr, _ = pearsonr(data[column], data[target_column])
        correlations[column] = corr
    return sorted(correlations.items(), key=lambda x: abs(x[1]), reverse=True)[:3]  # 取前3到4个最大绝对值的系数

top_features = calculate_pearson_coefficients(df, "光伏功率")

2. 确定重要特征并重新划分数据集：

selected_columns = [df.columns[i] for i, _ in top_features]
X_train_selected, X_test_selected, y_train, y_test = train_test_split(df[selected_columns], df["光伏功率"], test_size=0.2, random_state=42)

3. 使用重要特征重新构建模型：

reduced_model = LinearRegression()  # 或者尝试别的模型如Lasso、Ridge
reduced_model.fit(X_train_selected, y_train)

4. 重新评估模型：

y_pred_train_reduced = reduced_model.predict(X_train_selected)
score_train_reduced = r2_score(y_train, y_pred_train_reduced)
print(f"Training R^2 score with selected features: {score_train_reduced}")

y_pred_test_reduced = reduced_model.predict(X_test_selected)
score_test_reduced = r2_score(y_test, y_pred_test_reduced)
print(f"Testing R^2 score with selected features: {score_test_reduced}")

5. 绘制预测与实际值的对比图：

# 用新选出的特征绘制预测结果
plt.scatter(y_test, y_pred_test_reduced, alpha=0.5)
# ...（同上一步，使用new_predictions代替y_pred_test）