The Comparison between Multivariate Fuzzy Time Series and Traditional Time Series Modeling to Forecasting China Exports

The paper evaluates three Multivariate Fuzzy Time Series models (MFTS) and traditional times models for the comparison of forecasting accuracy. The data for model test are obtained from State Administration of Foreign Exchange Website, Mainland China, including monthly exports value and spot exchange rate from January 1995 to October 2002. The result indicates that China exports are influenced by the interfere variables of one-period ahead, twelve-period and thirteen-period. MFTS is more appropriate for the short-time prediction than traditional time series models. Furthermore, the one-variable MFTS model generally performs better forecasting accuracy than multi-variable model.