Performance Metrics
How to draw ROC?
In signal detection theory, a receiver operating characteristic (ROC), or simply ROC curve, is a graphical plot which illustrates the performance of a binary classifier system as its discrimination threshold is varied.
Most binary classifiers give a prediction probability for positive and negative classes. If you set a threshold say, 0.6, you will get a Recall(TPR) and False alarm(FPR). then you vary this threshold value, you will get a group of points. threshold value = 0 corresponds to the point (1,0) while threshold value = 1 corresponds to point(0,0)
[1] 机器学习之分类性能度量指标 : ROC曲线、AUC值、正确率、召回率
Why use ROC/AUC?
The AUC value is equivalent to the probability that a randomly chosen positive example i s ranked higher than a randomly chosen negative example. When data sets are imbalanced, ROC/AUC is more stable than Recall, F1, precision..
Confusion matrix
\