机器学习算法之决策树：随机森林（Random Forest）原理、手动计算与Python/Java双代码实战指南

from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import make_classification
from sklearn.inspection import permutation_importance

# 生成数据
X, y = make_classification(n_samples=1000, n_features=4, random_state=42)

# 训练
rf = RandomForestClassifier(
    n_estimators=100,
    max_features='sqrt',  # 特征随机
    oob_score=True,       # 启用OOB
    random_state=42
)
rf.fit(X, y)

# 预测 & OOB 误差
print("OOB Score:", rf.oob_score_)
print("预测:", rf.predict(X[:5]))

# 特征重要性（排列法）
perm_imp = permutation_importance(rf, X, y, n_repeats=10, random_state=42)
print("Permutation Importance:", perm_imp.importances_mean)

import numpy as np
from sklearn.tree import DecisionTreeClassifier

class SimpleRandomForest:
    def __init__(self, n_estimators=10, max_features='sqrt'):
        self.n_estimators = n_estimators
        self.max_features = max_features
        self.trees = []
        self.feature_indices = []

    def fit(self, X, y):
        n_samples, n_features = X.shape
        if self.max_features == 'sqrt':
            m = int(np.sqrt(n_features))
        else:
            m = n_features // 3

        for _ in range(self.n_estimators):
            # Bootstrap 采样
            idx = np.random.choice(n_samples, n_samples, replace=True)
            X_boot, y_boot = X[idx], y[idx]
            
            # 随机选特征
            feat_idx = np.random.choice(n_features, m, replace=False)
            self.feature_indices.append(feat_idx)
            
            # 训练树
            tree = DecisionTreeClassifier()
            tree.fit(X_boot[:, feat_idx], y_boot)
            self.trees.append(tree)

    def predict(self, X):
        predictions = []
        for i, tree in enumerate(self.trees):
            pred = tree.predict(X[:, self.feature_indices[i]])
            predictions.append(pred)
        # 多数投票
        return np.array([np.bincount(col).argmax() for col in np.array(predictions).T])

<dependency>
    <groupId>nz.ac.waikato.cms.weka</groupId>
    <artifactId>weka-stable</artifactId>
    <version>3.8.6</version>
</dependency>

import weka.classifiers.trees.RandomForest;
import weka.core.Instances;
import weka.core.converters.ConverterUtils.DataSource;

public class RandomForestExample {
    public static void main(String[] args) throws Exception {
        // 加载数据
        DataSource source = new DataSource("data.arff");
        Instances data = source.getDataSet();
        data.setClassIndex(data.numAttributes() - 1);

        // 配置随机森林
        RandomForest rf = new RandomForest();
        rf.setNumIterations(100);          // 树的数量
        rf.setNumFeatures(0);              // 0 = sqrt(numAttributes-1)
        rf.setComputeImportances(true);    // 计算特征重要性

        // 训练
        rf.buildClassifier(data);

        // 输出模型（含OOB误差和特征重要性）
        System.out.println(rf.toString());

        // 预测
        double pred = rf.classifyInstance(data.instance(0));
        System.out.println("预测类别: " + pred);
    }
}