ROC curve and AUC value

Link: https://www.zhihu.com/question/39840928/answer/146205830
Source: Zhihu

 

1. Confusion Matrix

The confusion matrix as shown in Figure 1 uses "0" and "1" to represent negative samples and positive samples, respectively. FP represents the number of samples for which the actual class label is "0", but the predicted class label is "1". The rest, similar reasoning.

 

2. False positive rate and true rate

False Positive Rate (FPR) is the proportion of samples with the actual label "0" that are predicted incorrectly. The True Positive Rate (TPR) is the proportion of the samples with the actual label "1" that are predicted to be correct. Its formula is as follows:

3. The ROC curve is a line connecting a series of (FPR, TPR) numerical points under the threshold. The value of threshold at this time is the predicted probability of each sample in the test data set.

AUC (Area Under roc Cure), as the name suggests, is the small area of ​​the ROC curve, in this example AUC=0.62. The larger the AUC, the better the classification effect.