zoukankan      html  css  js  c++  java
  • 交叉验证

    对于线性回归:
    方法一:以前的cross validation中有一种方法是train/test split,现在挪到model_selection库中,randomly partition the data into training and test sets, by default, 25 percent of the data is assigned to the test set。这种方法只能得到一次划分结果的评估结果,不准确。

    score算的是r-squared系数,好像score和cross_val_score默认算的就是r-squared系统

    // from sklearn.model_selection import train_test_split
    // X_train,X_test,y_train,y_test=train_test_split(X,y)
    // model=LinearRegression()
    // model.fit(X,y)
    // model.score(X_test,y_test)

    方法二:用model_selection库中的cross_val_score
    // from sklearn.model_selection import cross_val_score
    // model=LinearRegression()
    // scores=cross_val_score(model,X,y,cv=5)

    cv=5表示cross_val_score采用的是k-fold cross validation的方法,重复5次交叉验证

    实际上,cross_val_score可以用的方法有很多,如kFold, leave-one-out, ShuffleSplit等,举例而言:

    //cv=ShuffleSplit(n_splits=3,test_size=0.3,random_state=0)
    //cross_val_score(model, X,y, cv=cv)

    对于逻辑回归:
    逻辑回归用于处理分类问题,线性回归求解how far it was from the decision boundary(求距离)的评估方式明显不适合分类问题。
    The most common metrics are accuracy, precision, recall, F1 measure, true negatives, false positives and false negatives
    1、计算confusion matrix
    Confusion matrix 由 true positives, true negatives, false positives以及 false negatives组成。
    // confusion_matrix=confusion_matrix(y_test, y_pred)
    2、accuracy: measures a fraction of the classifier's predictions that are correct.
    // accuracy_score(y_true,y_pred)
    LogisticRegression.score() 默认使用accuracy
    3、precision: 比如说我们预测得了cancer中实际确实得病的百分比
    // classifier=LogisticRegression()
    // classifier.fit(X_train,y_train)
    // precisions= cross_val_score(classifier, X_train,y_train,cv=5,scoring='precision')
    4、recall: 比如说实际得了cancer,被我们预测出来的百分比
    // recalls= cross_val_score(classifier,X_train,y_train,cv=5,scoring='recall')
    5、precision和recall之间是一个trade-off的关系,用F1score来表征性能,F1score越高越好
    // fls=cross_val_score(classifier, X_train, y_train, cv=5,scoring='f1')
    6、ROC曲线和AUC的值
    ROC曲线的横坐标为false positive rate(FPR),纵坐标为true positive rate(TPR)
    AUC数值=ROC曲线下的面积
    // classifier=LogisticRegression()
    // classifier.fit(X_train, y_train)
    // predictions = classifier.predict_proba(X_test)
    // false_positive_rate, recall, thresholds = roc_curve(y_test, predictions[:,1])
    // roc_auc=auc(false_positive_rate, recall)



    作者:dechuan
    链接:https://www.jianshu.com/p/a4e94e72a46d
    來源:简书
    著作权归作者所有。商业转载请联系作者获得授权,非商业转载请注明出处。
  • 相关阅读:
    jquery图片播放弹出插件Fancybox
    D3js-API介绍【英】
    ZOJ 3156 Taxi (二分匹配+二分查找)
    linux权限之su和sudo的差别
    CareerCup之1.6 Rotate Image
    [oracle]pl/sql --分页过程demo
    已迁移到http://www.coffin5257.com
    C# 之 集合ArrayList
    Java 序列化Serializable具体解释(附具体样例)
    Android 短信验证码控件
  • 原文地址:https://www.cnblogs.com/zyy-/p/8530546.html
Copyright © 2011-2022 走看看