Construction of Fuzzy Classification System Based on Multi-objective Genetic Algorithm

This paper present a novel method based on multi-objective genetic algorithm to construct fuzzy classification system. First, in order to relieve the problem of "curse of dimensionality", a multi-objective genetic algorithm is used to accomplish feature selection and fuzzy partition with maximum classification performance and minimum number of features and minimum number of fuzzy rules, thus an initial fuzzy system is obtained. Then, a genetic algorithm is employed to select significant fuzzy rules with two objectives to achieve a compact fuzzy system. In order to improve the classification performance of the compact fuzzy system, a constrained genetic algorithm is utilized to optimize the parameters of the compact fuzzy system. The proposed approach is applied to the Iris and Wine benchmark problems, and the results show its validity.