應用多變量模糊時間序列預測台灣加權股價指數之研究

• Main Authors: Cheng-Kai Hung, 洪呈凱
• Other Authors: Mei-Ling Huang
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/78315966744141927012

碩士 === 國立勤益科技大學 === 工業工程與管理系 === 104 === Stock index which closely related to the national economy is an important indicator of the stock market. In recent years, global finance influenced the stock market in many changes, such as the Asian financial in A.D. 1997 and the global financial crisis in A.D. 2008. When government and corporation were absence of preparedness, the financial systems often faced crises, therefore the accuracy of the stock market forecast has become a major issue. The researches of the fuzzy time series have been proved that they performed well in forecasts than in the linear models. This study proposed multivariate fuzzy time series which utilized Vector Autoregression (VAR) implemented Ant Colony Optimization (ACO) to forecast the Taiwan weighted stock index. The data used 15 kinds of different stock indexes from A.D. 2000 to 2010, which included the Taiwan weighted stock index, NASDQ index, New York Stock index etc. for the forecasts. The VAR filtered the significant variants and selected the optimized lags, then used the ACO to build the optimized fuzzy rule table. The results showed that the proposed model was stable in forecasts, the coefficient of variation which was calculated by the 20 times repeatedly experiment was less than 0.1. Compared the accuracy with recently proposed fuzzy time series models, the proposed model improved 2% to 36% accuracy of forecasts.