Outliers detection and confidence interval modification in fuzzy regression

Abstract The existence of outliers in a set of experimental data can cause incorrect interpretation of the fuzzy linear regression results. Peters (Fuzzy Sets and Systems 63 (1994) 45–55) considered this problem for the nonfuzzy input and nonfuzzy output data type. The present investigation focuses on nonfuzzy input and fuzzy output data type and proposes approaches to handle the outlier problem. The main idea is to introduce a pre-assigned k -limiting value whose value must be determined based on the conditions of the current problem. The advantages and problems of the proposed approaches are compared and discussed.