This paper presents a method for addressing the issue of outliers in fuzzy data. The method involves calculating a new distance between fuzzy numbers using various kernel functions, based on the fuzzy least squares method. The parameters of the fuzzy regression model were estimated in cases where the explanatory variables were non-fuzzy, the parameters were fuzzy, and the response variable was both fuzzy and an outlier. These estimators were then compared to the Fuzzy Least Squares method (FLS) using Mean Square Error (MSE) through simulations with different sample sizes (25, 50, 100, 150) and levels of outliers (0, 0.10, 0.20, 0.30). The results showed that this method, utilizing the new distance, achieved the best results in the presence of outliers.                         

Paper type: Research paper.