16 Jan 2021

AUC presents threshold invariance (no need to set a threshold as in accuracy) and scale invariance (if you scale all predictions linearly, but maintain ranking, then resulting auc is the same), which necessarily means that the AUC metric doesnโ€™t depend on the chosen threshold or the scale of probabilities.

These properties make AUC pretty valuable for evaluating binary classifiers as it provides us with a way to compare them without caring about the classification threshold.

To plot a ROC (Receiver Operating Characteristic) curve, simply plot TPR as a function of FPR. That is,

true positives / total positives (recall) vs false positives / total negatives (how many errors did you make in the negative case)

It can also be mathematically proven that AUC is equal to the probability that a classifier will rank a randomly chosen positive instance higher than a randomly chosen negative one.

Guide to ROC Curves, Medium