### 使用 Python 实现随机森林预测并获取置信区间 为了实现这一目标，可以采用基于自助法（Bootstrap）的方法来构建多个随机森林模型，并利用这些模型的结果计算预测区间。下面介绍具体做法以及相应的代码示例。 #### 构建随机森林模型并应用自助法 通过多次抽样训练集数据建立不同的随机森林模型，对于每一个测试样本点，收集所有模型对该点的预测值形成分布，进而求得上下限作为预测区间[^1]。 ```python import numpy as np from sklearn.ensemble import RandomForestRegressor from sklearn.model_selection import train_test_split from sklearn.datasets import make_regression def bootstrap_predict(X_train, y_train, X_test, n_estimators=100, n_bootstrap_samples=100): predictions = [] for _ in range(n_bootstrap_samples): # Bootstrap sampling with replacement from training data indices = np.random.choice(len(X_train), size=len(X_train), replace=True) rf_model = RandomForestRegressor(n_estimators=n_estimators) rf_model.fit(X_train[indices], y_train[indices]) prediction = rf_model.predict(X_test).reshape(-1, 1) predictions.append(prediction) all_predictions = np.hstack(predictions) lower_bound = np.percentile(all_predictions, 2.5, axis=1) # Lower bound of the confidence interval (95% CI) upper_bound = np.percentile(all_predictions, 97.5, axis=1) # Upper bound of the confidence interval (95% CI) mean_prediction = np.mean(all_predictions, axis=1) # Mean prediction return mean_prediction, lower_bound, upper_bound # Example usage: X, y = make_regression(n_features=4, noise=0.2, random_state=1) X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42) mean_pred, lb, ub = bootstrap_predict(X_train, y_train, X_test) print("Mean Prediction:", mean_pred[:5]) print("Lower Bound:", lb[:5]) print("Upper Bound:", ub[:5]) ``` 此段代码展示了如何创建一个函数 `bootstrap_predict` 来执行上述过程，最终返回平均预测值及其对应的下界和上界的数组表示形式。这允许使用者直观地查看每个测试实例上的不确定性程度。