Multistep fuzzy classifier forming with cooperative-competitive coevolutionary algorithm

Abstract

This paper is about multistep fuzzy classifier forming method with cooperative-competitive coevolutionary algorithm. Cooperative-competitive coevolutionary algorithm automatically allows avoiding the problem of genetic algorithm parameters setting. This approach is included in a new method combining Michigan and Pittsburgh approaches for fuzzy classifier design. The procedure is performed several times. After each step classification efficiency is increased and standard deviation of values is decreased. Results of numerical experiments for machine learning problems from UCI repository are presented. © 2012 Springer-Verlag.