Relevance and redundancy in fuzzy classification systems

Abstract

Fuzzy classification systems is defined in this paper as an aggregative model, in such a way that Ruspini classical definition of fuzzy partition appears as a particular case. Once a basic {\em recursive} model has been accepted, we then propose to analyze relevance and redundancy in order to allow the possibility of {\em learning} from previous experiences. All these concepts are applied to a real picture, showing that our approach allows to check quality of such a classification system.