Multiple-Period Modified Fuzzy Time-Series for Forecasting TAIEX

In fuzzy time-series methods, mining fuzzy logical relation (FLR) from time-series is one of most critical processes to influence forecasting accuracy. However, in stock markets, investors usually make their investment decisions according to recent stock information such as market news, technical indicators or yesterday price. In this paper, we propose a new fuzzy time-series, which integrates linear relationships between recent periods of stock prices and non-linear relationships (FLR) in forecasting processes, to improve forecasting performance. To verify the proposed method, in this paper, we employ a nine-year period of TAIEX (Taiwan Stock Exchange Capitalization Weighted Stock Index) data as experimental dataset and three fuzzy time-series methods, Chen´s, Yu´s and Huarng´s methods, as comparison methods. The comparison results show that our method outperforms the listing methods which only consider FLR in forecasting processes.