Çoklu örnek öğrenimi için alıcı operatörü karakteristik eğrisi altında kalan alanı en iyileştiren kesin yöntem yaklaşımı

The purpose of this study is to solve the multi-instance classification problem by directly maximizing the area under Receiver Operating Characteristic (ROC) curve (i.e., AUC). We derive a mixed integer linear programming model that produces the best possible hyperplane-based classifier for multi-instance classification. Our study sheds a light on the potential of hyperplane-based approaches, reflecting cross validation (CV) results for benchmark instances. As we maximize AUC directly, a hyperplane-based classifier can only coincidentally provide a better CV accuracy than those presented in this paper. Finally, we present how Kernel trick can be applied to produce nonlinear classifiers that maximize AUC.