| Summary: | 碩士 === 國立嘉義大學 === 資訊工程學系研究所 === 105 === Stock market trend is a major concern for most investors, and therefore predicting stock market trend is an active research area for many years. Because financial news can easily affect the stock market and short-term trading is very popular in Taiwan, investors who can interpret the financial news correctly will have the advantage in predicting the stock price trend. With the rapid development of the media industry, investors can easily obtain the financial news but generally lack the time and ability to analyze the information. Therefore, having the ability to quickly and effectively analyze the large amount of news information is a key factor for making a profit from the stock market. At present, most studies on stock price prediction collect and analyze the financial news related to a specific stock, and then combine the news information with the historical stock price to build a forecast model. There are very few studies that have investigated how to use the integrated information from all the financial news of the day to predict the following day's changes of the Taiwan stock market. In this study, we use text mining and document classification techniques to analyze the financial news of the day, to extract useful feature information, and to predict the following day's ups and downs of the Taiwan stock market. We use different methods to combine all the financial news in a day into an appropriate feature set, which represents the day's news information. Experimental results show that the combined feature information can improve the accuracy of trend prediction of the Taiwan stock market.
|
|---|
