Interval type-2 fuzzy classifier design using Genetic Algorithms

Abstract

This paper aims at investigating the advantages of using an interval type-2 fuzzy system for classification problems. An evolutionary architecture was proposed to generate the rule base and to optimize the membership functions of a type-2 Fuzzy Classification System The proposed architecture is composed of three stages. In the first stage of the architecture, a Genetic Algorithm generates the rule base of the Fuzzy Classification System using predefined and fixed membership functions. In the second stage, another Genetic Algorithm optimizes the interval type-2 membership functions that were used in the first stage. Finally, a third Genetic Algorithm is used for the optimization of the number of rules in the best Fuzzy Classification System generated in the two previous stages. Some experiments have been run using different datasets from the UCI Machine Learning Repository in order to validate the proposed approach and to compare the results with the ones obtained with the Wang&Mendel method and a type-1 fuzzy classification system also generated by the evolutionary architecture proposed here. The results demonstrated that the type-2 fuzzy classification system performed better than the other classifiers used in the study.