auc and ranking - optimize your linear classifier to maximize AUC (and ranking). Close set ranking - make your classifier to provide ranks to your data, for example: "excellent", "good", "bad", "horrible" .