机器学习(周志华)课后题 第三章 线性模型 3.4

比较10折交叉验证法和留一法

加载函数包和数据集

from sklearn.model_selection import KFold,LeaveOneOut # 交叉验证所需的子集划分方法
from sklearn.linear_model.logistic import LogisticRegression
from sklearn import datasets  # 自带数据集
import numpy as np
iris = datasets.load_iris()  # 加载数据集
print('样本集大小：',iris.data.shape,iris.target.shape)

定义交叉验证函数

def partition(k, X, y, Model, Method):
    # Random seed: reproducibility
    np.random.seed(1)

    # accuracy score
    train_accuracy = [0 for i in range(k)]
    test_accuracy = [0 for i in range(k)]

    # index
    idx = 0

    if Method == 'KFold':#k折交叉验证法(k=10)
        m = KFold(n_splits=k, shuffle=True)
    if Method == 'LeaveOneOut':#留一法
        m = LeaveOneOut()

    # Generate the sets
    for train_index, test_index in m.split(X):
        # Iteration number
        # print(train_index,len(train_index))
        X_train, X_test = X[train_index], X[test_index]
        y_train, y_test = y[train_index], y[test_index]

        # Calling the function/model
        if Model == "Logit":
            clf = LogisticRegression(random_state=0)

        # Fit the model
        clf = clf.fit(X_train, y_train)
        y_train_pred = clf.predict(X_train)
        y_test_pred = clf.predict(X_test)

        train_accuracy[idx] = np.mean(y_train_pred == y_train)
        test_accuracy[idx] = np.mean(y_test_pred == y_test)
        idx += 1

    return train_accuracy, test_accuracy

进行对率回归

10折交叉验证法

train_acc,test_acc = partition(10,iris.data,iris.target,"Logit",'KFold')
print(np.mean(train_acc),np.mean(test_acc))

0.9607407407407406 0.9466666666666667

留一法

train_acc,test_acc = partition(len(iris.target),iris.data,iris.target,"Logit",'LeaveOneOut')
print(np.mean(train_acc),np.mean(test_acc))

0.9626845637583892 0.9533333333333334