Fuzzy regression model using fuzzy partition

Abstract

In this paper, we introduce a novel method to estimate fuzzy regression model when the center and spread have different pattern. We combine the least absolute deviations estimation for the center with discriminant analysis for spread. Also, we use the fuzzy partition to categorize spreads of the dependent variable into several classes. A numerical study is given to show that our method outperforms the estimation methods based on a fuzzy regression model with the assumption of equivalent functions for the center and spread.