Area under ROC

To assess the model/classifier, we need to determine the area under ROC (AUROC). The whole area of this plot is 1 as the maximum value of FPR and TPR – both are 1 here. Hence, it takes the shape of a square. The random line is positioned perfectly at 45 degrees, which partitions the whole area into two symmetrical and equilateral triangles. This means that the areas under and above the red line are 0.5. The best and perfect classifier will be the one that tries to attain the AUROC as 1. The higher the AUROC, the better the model is.

In a situation where you have got multiple classifiers, you can use AUROC to determine which is the best one among the lot.