A Comparison of Neural Network and the Decision Tree Approach to Predict Stock Performance

• Main Authors: Tsung-Yeng Yang, 楊宗彥
• Other Authors: Hsin-Cheng Tsen
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/se647c

碩士 === 逢甲大學 === 企業管理所 === 91 === Normally, the stock price should respond the true value of the company. The value of the company reflects the status of company’s management, cooperation condition and profitability. This paper tries to use the financial ratio of public company and two different mining methods to predict the future stock return. This study uses the electronic stocks in Taiwan from the first season of 1994 through the third season of 2002 as study samples, and all data information are collected from Taiwan Economic Journal (TEJ). The independent variables are the company’s financial ratios and the dependent variable is the stock return. This paper uses the neural networks and decision tree approach to predict stock return and try to compare the accuracy rate. Next, this paper tries to find out the changes of accuracy when the neural network and decision tree model get additional training data or variables. The conclusions of this paper are stated as following: A.Additional training data or variables shall be included to promote predict accuracy rate of stock return B.Using neural network tools to predict stock returns have higher accuracy rate than using decision tree approach. C.Increasing variables is better than increasing training data in two data mining tools.