An Application of Knowledge Discovery in Stock Market

• Main Author: 邱垂佳
• Other Authors: Shun-Feng Su
• Format: Others
• Language: en_US
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/23513130325503973029

碩士 === 國立臺灣科技大學 === 電機工程系 === 90 === In recent years, business and industry face with a flood of data everyday. In order to use those data further, the employment of “Knowledge Discovery in Database”(KDD) is necessary. Through the processes of KDD, make the raw data to become more significant for managers or users. Stock market is one kind of database that is full of various kinds of information. In general, the traditional stock analysis is to use technical indicators to deal with the data in the stock market. In this thesis, we attempt to employ the technique of KDD to deal with such a huge number of data in the stock market. The purpose of our study is to extract the relationship between two stocks. We use three methods to define the relationship and the delay days and then exploit the delay days to define an overall relationship between two stocks with a specified time window. When we gain the relationship and the delay days, three neural networks are used to find the correlations with traditional technical indicators. We train three neural networks with technical indicators as inputs, and the outputs are the relationship between two stocks, the delay days stock A lagging stock B, and the days stock B lagging stock A, respectively. According to the outcome of the neural networks, we can make prediction of the behavior of the stock market. Hopefully, the extracted knowledge can provide consultations for investors to adjust their strategy in the stock market.