The Fuzzy Logic Method for Simpler Forecasting

Abstract

Fildes and Makridakis (1998), Makridakis and Hibon (2000), and Fildes (2001) indicate that simple extrapolative forecasting methods that are robust forecast equally as well or better than more complicated methods, i.e. Box-Jenkins and other methods. We study the Direct Set Assignment (DSA) extrapolative forecasting method. The DSA method is a non-linear extrapolative forecasting method developed within the Mamdani Development Framework, and designed to mimic the architecture of a fuzzy logic control system. We combine the DSA method Winters' Exponential smoothing. This combination provides the best observed forecast accuracy in seven of nine subcategories of time series, and is the top three in terms of observed accuracy in two subcategories. Hence, fuzzy logic which is the basis of the DSA method often is the best method for forecasting.