今天给大家带来一篇如何评价模型的好坏以及模型的得分
最下面的代码最有用
一、错误率与精度(accuracy 准确)
错误率和精度是分类任务中最常用的两种性能度量,既适用于二分类任务,也适用于多分类任务。错误率是分类错误的样本数占样本总数的比例,精度则是分类正确的样本数占样本总数的比例。
from sklearn import metrics
print('模型精度:',metrics.accuracy_score(y_test,y_predict))
二、查准率(precision)、查全率(recall)与F1-score
查准率(P值)是针对我们的预测结果而言的,它表示的是预测为正的样本中有多少是真正的正样本
查全率(R值)是针对我们原来的样本而言的,它表示的是样本中的正例有多少被预测正确了
查准率 P与查全率 R 分别定义为
![](file:///C:/Users/%E4%BC%AA%E5%96%84/AppData/Local/Temp/enhtmlclip/20180722184250165.png)
![](data:image/png;base64,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)
查准率和查全率是一对矛盾的度量.一般来说,查准率高时,查全率往往偏低;而查全率高时,查准率往往偏低。
F1-score,是统计学中用来衡量二分类模型精确度的一种指标。它同时兼顾了分类模型的准确率和召回率。F1分数可以看作是模型准确率和召回率的一种加权平均,它的最大值是1,最小值是0。
随着阈值的变化,就像假设检验的两类错误一样,如下图所示召回率和精确率不能同时提高,因此我们就需要一个指标来调和这两个指标,于是人们就常用F1-score来进行表示:
![](file:///C:/Users/%E4%BC%AA%E5%96%84/AppData/Local/Temp/enhtmlclip/201807221908489.png)
![](data:image/png;base64,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)
print('查准率:',metrics.precision(y_test,y_predict))
print('查全率:',metrics.recall_score(y_test,y_predict))
print('F1-score:',metrics.precision_score(y_test,y_predict))
三、ROC曲线、AUC
ROC(Receiver Operating Characteristic) 受试者工作特征曲线的纵轴是"真正例率" (True Positive Rate,简称TPR) ,也称灵敏度,横轴是"假正例率" (False Positive Rate,简称FPR) ,也称1-特异度,两者分别定义为
print('AUC:',metrics.roc_auc_score(y_test,y_pred))
四、log-loss
很多机器学习的算法通常会用logloss作为模型评价的指标,对数损失(Log loss)亦被称为逻辑回归损失(Logistic regression loss)或交叉熵损失(Cross-entropy loss),简单来说就是逻辑回归的损失函数。
y_pred=LR.(predict_proba(X))[:,1]预测类别为1的概率
print('log-loss:',metrics.log_loss(y_test,y_pred))
#准确率(accuracy),精确(precision_weighted),召回(recall_weighted),F1(f1_weighted)
#导入评分的包
from sklearn.model_selection import cross_val_score
# cv=6 是把数据分成6分,交叉验证, mea平均数,确保数据的准确率
print('准确{}'.format(cross_val_score(gaussian,test_X,test_Y,scoring='accuracy',cv=6).mean()))
print('精确{}'.format(cross_val_score(gaussian,test_X,test_Y,scoring='precision_weighted',cv=6).mean()))
print('召回{}'.format(cross_val_score(gaussian,test_X,test_Y,scoring='recall_weighted',cv=6).mean()))
print('F1值{}'.format(cross_val_score(gaussian,test_X,test_Y,scoring='f1_weighted',cv=6).mean()))
#查看分类报告
from sklearn.metrics import classification_report
# 查看更详细的, target_names=iris.target_names()
print(classification_report(y_pred=y_pre,y_true = train_Y,target_names=iris.target_names))
#support :原个数类别个数
from sklearn import metrics
print('平均绝对误差:{}'.format(metrics.mean_squared_error(y_predict,trainY)))
print('均方误差MSE:{}'.format(metrics.mean_absolute_error(y_predict,trainY)))
print('解释方差分:{}'.format(metrics.explained_variance_score(y_predict,trainY)))
print('R2得分:{}'.format(metrics.r2_score(y_predict,trainY)))