Multiprototype-based fuzzy classification and reject options

This paper aims at presenting different classifiers (classification rules) with two characteristics. First, they are based on multiprototype fuzzy labels which are combined using connectives (t-conorms). Thus, the definition of the classes increases and consequently the classifier performance. Second, the rules include reject options. They allow the classifiers to manage uncertainty due to both imprecise and incomplete definition of the classes. Performance on artificial and real data are presented and discussed.