The Performance Comparisons between Error Correction Model and Neural Network－An Empirical Study in Taiwan

• Main Authors: Hsiang-mei Tseng, 曾湘嵋
• Other Authors: Mei-se Chien
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/92338947657897131589

碩士 === 國立高雄應用科技大學 === 金融資訊研究所 === 93 === The investment has become prevailing in Taiwan in recent year. By investing on all kinds of financial products, stock is the most popular investment for participants. In order to get substantial returns, people should master market information in the fluctuating stock market. For the reasons, how to predict the stock price correctly becomes the topic issue. From the views of macroeconomics, this study used three methods, including econometrics method, neural network and synthetic of these two methods, to construct empirical models for predicting the stock index in Taiwan, and then compared the performance of these different empirical models. Our major findings from empirical research were as follows: First, the performance of the error correct model was better than other econometrics methods. Second, the performance of the synthetic of backpropagation neural networks and econometrics methods was better than the traditional backpropagation neural networks. The result also showed if we have used econometrics method to modify input and output of the empirical data, it could really improve the predicting performance of traditional backpropagation neural networks. Third, the performance of the synthetic of backpropagation neural networks and econometrics methods were indeed better than the traditional econometrics methods and backpropagation neural network. Overall, the error correct model and the synthetic of backpropagation neural networks and econometrics methods can predict the long-term trend of stock index in Taiwan. Consequently, these two kinds of the empirical models were relatively suitable tools for analyzing the long-term trend of stock market. Key words： Stock index、Macroeconomics、Error correct model、Neural networks