Forecasting TAIWAN Stock Markets with Technical Indicators: A Comparison between Neural Network Model and Multiple Regression Model

• Main Authors: Wang,Chun-Sheng, 王春笙
• Other Authors: Lee,Anthony J.T
• Format: Others
• Language: zh-TW
• Published: 1996
• Online Access: http://ndltd.ncl.edu.tw/handle/18057332331002908481

碩士 === 國立臺灣大學 === 資訊管理研究所 === 84 === With the progress of information technology, it is more and more popular to use IT in the real life. This study is to forecast TAIWAN stock market. Two methods are used in this study; one is back-propagation neural network, the other is statistical multiple regression. In addition, two decision support systems are also designed to trade stocks automatically. There are 18 companies sampled in this study. Ten out of 32 technical indicators are selected. These 10 indicators are fixed inputs of all neural network and statistical models. Each method has three kinds of models. One is to forecast the fluctuation of stock prices in 6 days, one is to forecast the fluctuation in 12 days, and the last is to forecast the fluctuation in 18 days. Using these models, this study computes the accuracy rate of forecasting and performance of trading simulation. Results reveal that one can forecast TAIWAN stock prices with certain accuracy. Back-propagation is better than multiple regression which is better than buy-and- hold strategy. The most profitable model in this study is to trade stock gradually by neural network model using the 18-days- later prices as target variables.