本文主要提供源码的一些思路.具体源码可查看知乎:知乎
简单来说随机森林就是生成N颗CART树,通过bootstrap的方式,有放回可重复的从原始数据集M里选出一部分数据m,总共生成N份这样的数据给N颗CART树去做训练,同时设定每棵树选用数据集中的最大的特征数,也是可重复的选取.最后的结果通过投票表决决定最终结果.
代码如下(没有运行,只是看了下思路):

# -*- coding: utf-8 -*-

import numpy as np
from decision_tree_model import ClassificationTree
from sklearn import datasets
from sklearn.model_selection import train_test_split

class RandomForest():
    """Random Forest classifier. Uses a collection of classification trees that
    trains on random subsets of the data using a random subsets of the features.
    Parameters:
    -----------
    n_estimators: int
        树的数量
        The number of classification trees that are used.
    max_features: int
        每棵树选用数据集中的最大的特征数
        The maximum number of features that the classification trees are allowed to
        use.
    min_samples_split: int
        每棵树中最小的分割数，比如 min_samples_split = 2表示树切到还剩下两个数据集时就停止
        The minimum number of samples needed to make a split when building a tree.
    min_gain: float
        每棵树切到小于min_gain后停止
        The minimum impurity required to split the tree further.
    max_depth: int
        每棵树的最大层数
        The maximum depth of a tree.
    """

    def __init__(self, n_estimators=100, min_samples_split=2, min_gain=0,
                 max_depth=float("inf"), max_features=None):

        self.n_estimators = n_estimators #树的数量
        self.min_samples_split = min_samples_split #每棵树中最小的分割数，比如 min_samples_split = 2表示树切到还剩下两个数据集时就停止
        self.min_gain = min_gain   #每棵树切到小于min_gain后停止
        self.max_depth = max_depth  #每棵树的最大层数
        self.max_features = max_features #每棵树选用数据集中的最大的特征数

        self.trees = []
        # 建立森林(bulid forest)
        for _ in range(self.n_estimators):
            tree = ClassificationTree(min_samples_split=self.min_samples_split, min_impurity=self.min_gain,
                                      max_depth=self.max_depth)
            self.trees.append(tree)

    def fit(self, X, Y):
        # 训练，每棵树使用随机的数据集(bootstrap)和随机的特征
        # every tree use random data set(bootstrap) and random feature
        sub_sets = self.get_bootstrap_data(X, Y)
        n_features = X.shape[1]

        if self.max_features == None:
            self.max_features = int(np.sqrt(n_features))
        for i in range(self.n_estimators):
            # 生成随机的特征
            # get random feature
            sub_X, sub_Y = sub_sets[i]
            idx = np.random.choice(n_features, self.max_features, replace=True)
            sub_X = sub_X[:, idx]
            self.trees[i].fit(sub_X, sub_Y)
            self.trees[i].feature_indices= idx
            print("tree", i, "fit complete")

    def predict(self, X):
        y_preds = []
        for i in range(self.n_estimators):
            idx = self.trees[i].feature_indices
            sub_X = X[:, idx]
            y_pre = self.trees[i].predict(sub_X)
            y_preds.append(y_pre)
        y_preds = np.array(y_preds).T
        y_pred = []
        for y_p in y_preds:
            # np.bincount()可以统计每个索引出现的次数
            # np.argmax()可以返回数组中最大值的索引
            # cheak np.bincount() and np.argmax() in numpy Docs
            y_pred.append(np.bincount(y_p.astype('int')).argmax())
        return y_pred

    def get_bootstrap_data(self, X, Y):

        # 通过bootstrap的方式获得n_estimators组数据
        # get int(n_estimators) datas by bootstrap

        m = X.shape[0] #行数
        Y = Y.reshape(m, 1)

        # 合并X和Y，方便bootstrap (conbine X and Y)
        X_Y = np.hstack((X, Y)) #np.vstack():在竖直方向上堆叠/np.hstack():在水平方向上平铺
        np.random.shuffle(X_Y) #随机打乱

        data_sets = []
        for _ in range(self.n_estimators):
            idm = np.random.choice(m, m, replace=True) #在range(m)中,有重复的选取 m个数字
            bootstrap_X_Y = X_Y[idm, :]
            bootstrap_X = bootstrap_X_Y[:, :-1]
            bootstrap_Y = bootstrap_X_Y[:, -1:]
            data_sets.append([bootstrap_X, bootstrap_Y])
        return data_sets


if __name__ == '__main__':
    data = datasets.load_digits()
    X = data.data
    y = data.target

    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, random_state=2)
    print("X_train.shape:", X_train.shape)
    print("Y_train.shape:", y_train.shape)

    clf = RandomForest(n_estimators=100)
    clf.fit(X_train, y_train)
    y_pred = clf.predict(X_test)

    accuracy = accuracy_score(y_test, y_pred)
    print('y_test:{}\ty_pred:{}'.format(y_test, y_pred))
    print("Accuracy:", accuracy)