Nonlinear pattern classification using fuzzy system and hybrid genetic-imperialist algorithm

Abstract

An approach proposed in this paper allows to select neuro-fuzzy classifiers taking into account new interpretability criteria. Those criteria are focused not only on complexity of the system, but also on semantics of the rules. The approach uses capabilities of new hybrid population algorithm which is a combination of the genetic algorithm and the imperialist competitive algorithm. This combination allows to select not only the parameters of the neuro-fuzzy system, but also the structure of it. In simulations typical issues of classification were used.