| Summary: | 碩士 === 元智大學 === 資訊管理學系 === 106 === In the stock market, news messages are the activities that deliver the fastest responses. Although the stock market operation generally gains profits from rational operations, how many people can truly follow the rational principles and make profits. Investors typically eager to make profits in a short time based on news information. This study aims to extract financial events from news articles such as Chinatimes from 2013 to 2017 as text sources. For each sentence, the verbs in the form of compound words are selected as financial events, and the likelihood of occurrence of financial events is then calculated as a screening condition. Finally, the average return rate and standard deviation are calculated for each event within 10 trading days.
To further investigate the prediction ability of machine learning algorithms for stock price movements, this study uses several structural features such as opening price, closing price and volume to train a random forest, logistic regression and support vector machine model for stock price prediction. Experimental results show that the upport vector machine achieves the best prediction performance.
|
|---|
