ROC

In general, predictor models output continuous prediction score. For example, in binary classification, it will be between 0 and 1. The value closer to 1 means prediction outcome positive, and the value closer to 0 means prediction outcome negative. In this case, we need a threshold as the prediction boundary. This threshold has a significant impact on all the performance metric we have discussed so far.
 
Receiver operating characteristic, or ROC curve, provide a way to compare different classifiers as a prediction boundary is varied, and this curve is created by plotting the true positive rate against the false positive rate at various threshold values. Such a curve can be generated by going through all the data points in the descending order of their prediction score and using those prediction scores as threshold values.

Видео ROC канала Precision Health