A COMPARISON OF FUZZY AND NONPARAMETRIC LINEAR-REGRESSION

Nonparametric linear regression and fuzzy linear regression have been developed based on different perspectives and assumptions, and thus th ere exist conceptual and methodological differences between the two ap proaches. This article describes their comparative characteristics suc h as basic assumptions, parameter estimation, and applications, and th en compares their predictive and descriptive performances by a simulat ion experiment to identify the conditions under which one method perfo rms better than the other. The experimental results indicate that nonp arametric linear regression is superior to fuzzy linear regression in predictive capability, whereas their descriptive capabilities depend o n various factors. When the size of the data set is small, error terms have small variability, or when the relationships among variables are not well specified, fuzzy linear regression outperforms nonparametric linear regression with respect to descriptive capability The conditio ns under which each method can be used as a viable alternative to the conventional least squares regression are also identified. The finding s of this article would be useful in selecting the proper regression m ethodology to employ under specific conditions for descriptive and pre dictive purposes. (C) 1997 Elsevier Science Ltd.