Multi-Strategy Multimodal Genetic Algorithm for Designing Fuzzy Rule Based Classifiers

Abstract

A hybridization of genetic algorithms and machine learning techniques have proved its effectiveness for many complex benchmark and real-world problems. In this study we present a novel approach that combines self-configuring genetic algorithm for multimodal optimization and fuzzy rule based classifier. The proposed search metaheuristic controls the interactions of many techniques for multimodal optimization (different genetic algorithms) and leads to the self-configuring solving of problems with a priori unknown structure. Appling this approach to designing the fuzzy rule based classifiers, we can obtain many optimal solutions with different representation. The results of numerical experiments with popular optimization benchmark problems (for multimodal genetic algorithm) and with well-studied real-world classification problems (for self-configuring fuzzy rule based classifier design) are presented and discussed. The main feature of the proposed approach is that it does not require the participation of the human expert, because it operates in an automated, self-configuring way.