When should I use The Area Under an ROC Curve (AUC) or the Confusion Matrix for classifier evaluation? The clasifier evaluation is for example the prediction of customers for possible future sales.
A confusion matrix can be used to measure the performance of a particular classifier with a fixed threshold. Given a set of input cases, the classifier scores each one, and score above the threshold are labelled Class 1 and scores below the threshold are labelled Class 2.
The ROC curve, on the other hand, examines the performance of a classifier without fixing the threshold. Given a set of input cases, the classifier scores each one. The ROC curve is then generated by testing every possible threshold and plotting each result as a point on the curve.
The ROC curve is useful when you want to test your classifier over a range of sensitivities/specificities. This may or may not be a desirable thing to do. Perhaps you want very high sensitivity and don't care much about specificity - in this case, the AUC metric will be less desirable, because it will take into account thresholds with high specificity. The confusion matrix, on the other hand, could be generated with a fixed threshold known to yield high sensitivity, and would only be evaluated for that particular threshold.
A confusion matrix evaluates one particular classifier with a fixed threshold, while the AUC evaluates that classifier over all possible thresholds.