Apply Gray Prediction and Evolutional Neural Network to Construct Taiwan Weighted Stock Index Forecast Models

• Main Authors: Hui-Chuan Wang, 王惠娟
• Other Authors: Tsung-Nan Chou
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/y8245a

碩士 === 朝陽科技大學 === 財務金融系碩士班 === 93 === Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX index) is the underlying of Dow Jones Taiwan Stock Index Futures, MSCI Taiwan Index Futures, and TAIEX Futures. It is an index of considerable weight when stock or futures investors make investment decisions. The stock market is susceptible to the influence of artificial, economic and political factors. Thus predicting the movement of stock prices is a challenging task. Moreover, stock market is a dynamic and open market. Any trading strategy might deviate off the course under vacillating environment and the influence of myriad factors. Thus any analytical theory must possess self-adjusting ability to adapt to the ever-changing market environment. This study aims to utilize the search power of evolutionary artificial neural networks and the predictive power of grey forecasting to predict the closing level and the movement of TAIEX index. The evolutionary artificial neural network model is used to predict the closing index using the day’s source data of TAIEX index, spread between TAIEX futures and spot, and the rise/fall of Dow Jones Industry Average Index (DJ) and NASDAQ index on previous closing as input parameters. In grey forecasting, a four-point rolling over grey model GM (1,1) is used to predict the closing TAIEX index and its direction of movement. The study also compares and analyzes the empirical results of these two different forecast models. Both models used in the study have their merits and theoretical basis. But which one is more suitable for predicting the closing level and movement of TAIEX index is also a point of discussion in the study. The empirical results of the two models find that the predictive power of grading-adjusted grey model GM(1,1) is markedly superior to that of evoluationary neural network model. It is also found that increasing the number of input parameter of neural network model does not necessarily improve its predictive power, whereas taking out the parameters of rise/fall of Dow Jones Industry Average Index (DJ) and NASDAQ index on previous closing achieved the best prediction results, indicating the less-than-expected influence of “overnight return” information on TAIEX that perhaps “latest information” on major events occurred prior to market opening should be taken into consideration. In forecasting the up/down movement of TAIEX index, the combination of two models resulted in the highest accuracy rate, while grey model has better predictive power than unadjusted model. The grey forecasting model may be further applied in other fields and the prediction of derivative or individual stock prices. It may also be used to gain understanding of investor’s reactions towards certain information.