A neural stock price predictor using qualitative and quantitative data

Most existing financial prediction systems are purely based on quantitative data such as stock prices and market indices. Qualitative information, such as company news or newsgroup discussions, are handled by the financial investor and not by any intelligent financial tools. As a result, most investors suffer from qualitative information overload and could miss an important shift in market sentiment. In this paper, a semi-automated approach to stock price prediction using newsgroup and financial information is discussed. The examples used in this paper are from the Singapore financial market. Singapore Stock Newsgroup articles from September 1996 to May 1997 were processed and the results showed that newsgroup articles have influenced stock price movements. Similarly, historical stock prices of a few major component stocks on the Singapore Stock Exchange (SES) from January 1996 to June 1997 were also obtained, and a time-delay neural network (TDNN) was used for the prediction of stock prices. Various ways to integrate the newsgroup articles and historical stock prices were then explored in order to try to generate a better prediction