Fuzzy time series models for LNSZZS forecasting

Abstract

Fuzzy time series methods have been recently becoming very popular in forecasting, it has been applied to forecast various domain problems and have been shown to forecast better than other models. But there still some suspicion whether fuzzy time series models is assured to better than ARMA, which is the best defuzzifying method. The purpose of this paper is to answer these two questions. In this paper, we use traditional the fuzzy time series method purposed by Song and Chissom to forecast SSE Composite Index (LNSZZS). In order to compare the forecast results, we used Maximum degree of membership method and Degree of membership weighted average method for defuzzifying. After forecasting by fuzzy time series method, we also compared the result to that calculated by ARMA. The research found that fuzzy time series model is more accurate than ARMA. We also see that for fuzzy time series forecasting model in defuzzifying, maximum degree of membership method is better than Degree of membership weighted average method.