Fuzzy Time Series Models Based on Fitting Function for Forecasting Stock Index

• Main Authors: Wei-Lun Tsai, 蔡維倫
• Other Authors: Cheng Ching-Hsu
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/35714909641059730760

碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 100 === In the recent years, many time series model has been widely applied in forecasting stock index. However, the time series methods still have some problem as follows: (1) conventional time series models only considered single variable; (2) fuzzy time series model determined the interval length of linguistic value subjectively; (3) selecting variables depended on personal experience and opinion. Hence, this paper proposes a novel fuzzy time series model based on fitting function to forecast stock index. The proposed model employed Pearson’s correlation to select important technical indicators objectively. In order to evaluate the performance of the proposed model, the transaction records of TAIEX (Taiwan Stock Exchange Capitalization Weighted Stock index) and HSI (Hang Seng Indexes) from 1998/01/03 to 2006/12/31 are used as experimental dataset and the root mean square error (RMSE) as evaluation criterion. And Chen’s (2000) model, Yu’s (2005) model support vector regression (SVR) and partial least square regression (PLSR) are used as comparable models with our methods The results show that the proposed model outperforms the listing models in accuracy for forecasting Taiwan stock market and Hong Kong stock market.