An Evaluation of Comparison between Multivariate Fuzzy Time Series with Traditional Time Series Model for Forecasting Taiwan Export

Abstract

Fuzzy time series methods have been applied for forecasting problems for over one decade. In this paper, we evaluate the forecasting accuracy by comparing three popular multivariate fuzzy time series models (MFTS) with two traditional time series models. The real world case of Taiwan exports is employed for modelspsila test to compare the forecasting ability among models and to examine the effects of different lengths of interval and increment information on the forecasting error of models. The data used for modelpsilas test includes two factors obtained from AREMOS, Taiwan. The results illustrates that MFTS are more appropriate for a short term prediction than ARIMA. Introducing increment information is not necessarily in improving the forecasting ability of fuzzy time series. Moreover, Heuristic method has the lowest MSE in MFTS.