sklearn K折(KFold)交叉验证案例，展开细节代码与cross_validate简写

一、通常的随机森林模型代码

对于一个基本的随机森林预测模型：

from sklearn.ensemble import RandomForestClassifier
import pandas as pd
import numpy as np
from sklearn.metrics import accuracy_score


def get_train_x_y():
    from sklearn.preprocessing import StandardScaler
    x = pd.DataFrame(data=np.random.randint(0, 10, size=(1000, 5)))
    y = np.random.randint(0, 2, 1000)
    x = StandardScaler().fit_transform(x)
    return x, y


if __name__ == '__main__':
    x_train, y_train = get_train_x_y()
    # 切分训练集与测试集
    from sklearn.model_selection import train_test_split

    x_train, x_test, y_train, y_test = train_test_split(x_train, y_train, test_size=0.3)
    # 使用随机森林预测
    my_model = RandomForestClassifier(n_estimators=10)
    my_model.fit(x_train, y_train)
    # 预测并得到准确率
    result_prediction = my_model.predict(x_test)
    score = accuracy_score(y_test, result_prediction)
    print(score)  # 得到预测结果区间[0,1]

二、K折交叉验证的随机森林代码

1. 切分方式：随机切分

使用K折前，前面都不变，从main函数开始：

if __name__ == '__main__':
    x_train, y_train = get_train_x_y()
    # 切分训练集与测试集，注意所有的交叉验证等都是在训练集上做的操作，测试集只有最后的最后才会使用到
    from sklearn.model_selection import train_test_split

    x_train, x_test, y_train, y_test = train_test_split(x_train, y_train, test_size=0.3)
    # 使用随机森林建模
    my_model = RandomForestClassifier(n_estimators=10)

然后就是设置KFold，n_splits=5时，切分训练集为5组

    # 使用交叉验证
    from sklearn.model_selection import KFold

    kfold = KFold(n_splits=5)
    
    for train_index, test_index in kfold.split(x_train, y_train):
        # train_index 就是分类的训练集的下标，test_index 就是分配的验证集的下标
        this_train_x, this_train_y = x_train[train_index], y_train[train_index]  # 本组训练集
        this_test_x, this_test_y = x_train[test_index], y_train[test_index]  # 本组验证集
        # 训练本组的数据，并计算准确率
        my_model.fit(this_train_x, this_train_y)
        prediction = my_model.predict(this_test_x)
        score = accuracy_score(this_test_y, prediction)
        print(score)  # 得到预测结果区间[0,1]

之后可以把每一组的score都保存起来，计算平均方差什么的。

KFold更多参数请参考：https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.KFold.html

2.切分方式：不均衡数据集下按比例切分

当然切分可以选择不同的切分方式，比如不均衡的数据集：

from sklearn.model_selection import StratifiedKFold

kfold = StratifiedKFold(n_splits=5)

这个会先将y分层，然后按照与训练集同样的比例进行切分

更多参数请参考：https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedKFold.html

三、KFold的简便写法

不需要把kfold切分开，只需要用一个cross_validate工具就行：

    # 使用交叉验证
    from sklearn.model_selection import KFold
    kfold = KFold(n_splits=5)

    # 使用K-fold的简写方式
    from sklearn.model_selection import cross_validate

	# 参数依次是：模型，训练集x，训练集y，kfold，评价指标
    cv_cross = cross_validate(my_model, x_train, y_train, cv=kfold, scoring=('accuracy', 'f1'))

cv_cross会返回每一轮的训练与预测时间，还有评价指标值，可以大大减少代码量

更多scoring请参考：https://scikit-learn.org/stable/modules/model_evaluation.html#the-scoring-parameter-defining-model-evaluation-rules

四、随机森林预测与KFold交叉验证完整代码

from sklearn.ensemble import RandomForestClassifier
import pandas as pd
import numpy as np


def get_train_x_y():
    from sklearn.preprocessing import StandardScaler
    x = pd.DataFrame(data=np.random.randint(0, 10, size=(1000, 5)))
    y = np.random.randint(0, 2, 1000)
    x = StandardScaler().fit_transform(x)
    return x, y


if __name__ == '__main__':
    x_train, y_train = get_train_x_y()
    # 切分训练集与测试集，注意所有的交叉验证等都是在训练集上做的操作，测试集只有最后的最后才会使用到
    from sklearn.model_selection import train_test_split

    x_train, x_test, y_train, y_test = train_test_split(x_train, y_train, test_size=0.3)
    # 使用随机森林建模
    my_model = RandomForestClassifier(n_estimators=10)
    # 使用交叉验证
    from sklearn.model_selection import KFold

    kfold = KFold(n_splits=5)

    # 使用K-fold的简写方式
    from sklearn.model_selection import cross_validate

    cv_cross = cross_validate(my_model, x_train, y_train, cv=kfold, scoring=('accuracy', 'f1'))