Improving the genetic optimization of fuzzy rule base by imposing a constraint condition on the number of rules

Pablo A. D. CastroHeloisa A. Camargo

A genetic algorithm-based learning procedure for designing fuzzy classification rules was proposed earlier by the authors and the results, advantages and usefulness of the proposal have been reported in the literature. The procedure is based on the Pittsburgh approach and is divided in two separate phases: genetic learning of candidate rules from numerical data and genetic optimization of previous obtained rules. Although the methodology presents very good results regarding accuracy and performance, the size of the obtained rules bases can be improved further. When compared with other approaches in the literature, the rules bases generated by our previous approach presented higher number of rules in most of the cases studied. In this work we propose a modification in the fitness function of the second stage imposing a constraint condition on the number of rules. Through computer simulations, we demonstrate that such a modified fitness function improves the efficiency of our approach, generating even more compact fuzzy rules bases.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: