Transparent fuzzy modeling based on minimum cluster volume

In this paper, a new method is proposed for transparent fuzzy modeling, which involves the minimum cluster volume (MCV) clustering algorithm. This method enables us to obtain membership functions with less overlap and larger core regions, then both ordinary least square and total least square methods are employed for less bias estimation of the consequent parameters. The characteristic of MCV applied to fuzzy modeling is analyzed from the perspectives of fuzzy modeling, such as the robustness of MCV, the cost function which contributes to simplify the rule base, etc. The modeling result gives easier understanding about the system than other clustering methods, such as available information for linear or nonlinear properties of the system, more accurate local model of the system and linguistic understanding of the system.