ROC curve

“Receiver operating characteristic” curve. A common way to examine the performance of a binary classifier. See also Class imbalance

• https://en.wikipedia.org/wiki/Receiver_operating_characteristic

One way to think about the ROC curve is considering the Decision theory. If a false prediction is $C$ times more costly than the correct prediction and if we draw the decision boundary accordingly, what would be the accuracy of the model?

If the ROC curve of a model is above another, the model dominates the other; regardless of $C$, the model is always better than the other.

Cross validated: Why are the ROC curves not smooth?