Classifiers are commonly evaluated using either a numeric metric, such as accuracy, or a graphical representation of performance, such as a receiver operating characteristic (ROC) curve.