GA search for fuzzy models under multiple-criteria

This paper presents a new framework for fuzzy modeling using genetic algorithm. A model of an actual object in the real world should satisfy various criteria, such as precision, generality and noise immunity. It is difficult for fuzzy modeling to allocate proper weights on these criteria. The framework introduced in this paper consists of a model generation block and a model testing block. The model generation block generates candidates of fuzzy models under criteria with higher importance, and the model testing block tests the candidates under not-so-important criteria. This division of criteria can put emphasis on the criteria in the generation block and less on those in the testing block. Simulations are done to show the effectiveness of the proposed framework.