sklearn中cross_val_score、cross_val_predict的用法比较

素颜马尾好姑娘i 2023-07-13 12:29 110阅读 0赞

交叉验证的概念，直接粘贴[scikit-learn官网][scikit-learn]的定义：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA_size_16_color_FFFFFF_t_70][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA_size_16_color_FFFFFF_t_70 1][]

--------------------

**scikit-learn中计算交叉验证的函数：**

cross\_val\_score：得到K折验证中每一折的得分，K个得分取平均值就是模型的平均性能

cross\_val\_predict：得到经过K折交叉验证计算得到的每个训练验证的输出预测

--------------------

**方法：**

cross\_val\_score：分别在K-1折上训练模型，在余下的1折上验证模型，并保存余下1折中的预测得分

cross\_val\_predict：分别在K-1上训练模型，在余下的1折上验证模型，并将余下1折中样本的预测输出作为最终输出结果的一部分

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA_size_16_color_FFFFFF_t_70 2][]

**结论：**

cross\_val\_score计算得到的平均性能可以作为模型的泛化性能参考

cross\_val\_predict计算得到的样本预测输出不能作为模型的泛化性能参考

--------------------

代码样例：

from sklearn import datasets
    import numpy as np
    from sklearn.tree import DecisionTreeClassifier
    from sklearn import datasets
    import numpy as np
    from sklearn.tree import DecisionTreeClassifier
    
    # 加载鸢尾花数据集
    iris = datasets.load_iris()
    iris_train = iris.data
    iris_target = iris.target
    print(iris_train.shape)
    print(iris_target.shape)
    (150, 4)
    (150,)
    
    # 构建决策树分类模型
    tree_clf = DecisionTreeClassifier()
    tree_clf.fit(iris_train, iris_target)
    tree_predict = tree_clf.predict(iris_train)
    
    # 计算决策树分类模型的准确率
    from sklearn.metrics import accuracy_score
    print("Accuracy:", accuracy_score(iris_target, tree_predict))
    Accuracy: 1.0
    
    # 交叉验证cross_val_score输出每一折上的准确率
    from sklearn.model_selection import cross_val_predict, cross_val_score, cross_validate
    tree_scores = cross_val_score(tree_clf, iris_train, iris_target, cv=3)
    print(tree_scores)
    [0.98039216 0.92156863 1.        ]
    
    # 交叉验证cross_val_predict输出每个样本的预测结果
    tree_predict = cross_val_predict(tree_clf, iris_train, iris_target, cv=3)
    print(tree_predict)
    print(len(tree_predict))
    print(accuracy_score(iris_target, tree_predict))
    [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
     0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1
     1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 1 2 2 2 2
     2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 1 2 2 2 2 2 2 2 2 2
     2 2]
    150
    0.96
    
    print(tree_clf.predict(iris_train))
    [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
     0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
     1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2
     2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
     2 2]
    
    # 交叉验证cross_validate对cross_val_score结果进行包装，并包含fit的时间等信息
    tree_val = cross_validate(tree_clf, iris_train, iris_target, cv=3)
    print(tree_val)
    {'fit_time': array([0., 0., 0.]), 'score_time': array([0., 0., 0.]), 'test_score': array([0.98039216, 0.92156863, 0.97916667])}

交叉验证评价方式scoring的参数链接：[https://scikit-learn.org/stable/modules/model\_evaluation.html][https_scikit-learn.org_stable_modules_model_evaluation.html]

[scikit-learn]: https://scikit-learn.org/stable/modules/cross_validation.html
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20200312143038791.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA_size_16_color_FFFFFF_t_70 1]: https://img-blog.csdnimg.cn/20200312143156767.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA==,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA_size_16_color_FFFFFF_t_70 2]: https://img-blog.csdnimg.cn/20200312105530934.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2N4eDY1NA==,size_16,color_FFFFFF,t_70
[https_scikit-learn.org_stable_modules_model_evaluation.html]: https://scikit-learn.org/stable/modules/model_evaluation.html