A hybrid approach to develop an interval type-2 fuzzy logic system

After more than three decades since the introduction of linguistic variables and their application to approximate reasoning by Zadeh [1], the ability of fuzzy logic systems (FLSs) for modeling real world applications is not a secret to anyone. Currently there are two basic approaches to determine fuzzy model of a system in the literature which are, 1-direct approach, and 2-indirect approach. In direct approach rules are generated via knowledge extraction from experienced experts, while in indirect approach historical data of a system determine the governing rules. The first method is involved with extracting knowledge from experts who in some cases are not available, or they avoid providing us with useful information. In the second method which is dealt with historical data, clustering is the proper tool for structure identification of a system under investigation. Determining the structure of a system relying only on past data also has its own problems. In this paper we try to develop a hybrid approach in interval type-2 fuzzy system modeling (IT2FSM) which benefits from the advantages of both direct and indirect methods. At first stage the modified approach to interval type-2 fuzzy c-mean clustering (IT2FCM) is applied to identify the structure of system and in the second stage the hybrid of direct and indirect approach in system modeling is used to complete the rule base of a model.