A proposal for an intuitionistic fuzzy inference system

Abstract

This work describes a method to construct type-1 intuitionistic fuzzy inference systems. This type of systems is able to handle more uncertainty than a type-1 fuzzy inference system and performs faster than a type-2 fuzzy inference system. The concepts of intuitionistic membership, and intuitionistic center of area are proposed, in order to implement a system which is similar in design than the traditional fuzzy inference systems. The proposed method was implemented and compared against type-1 fuzzy inference systems and interval type-2 fuzzy inference systems, with uncertain means and uncertain standard deviations, by using the Mackey-Glass time series benchmark. A genetic algorithm was used to optimize the parameters of each of the methods being compared. This optimization ensures a fair comparison between each of the methods against the proposed method. The results show that the intuitionistic fuzzy inference system performs better than the other methods.