Applying Deep Learning to Study the Correlation between Internet Public Opinions and the Movements of Stock Price — PTT BBS as an Example

• Main Authors: WANG, YOU-JIE, 王宥杰
• Other Authors: CHENG, LI-CHEN
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/y2pc3s

碩士 === 東吳大學 === 資訊管理學系 === 106 === The study observe how the public opinion of internet impacts the movement of the stock price, which uses the posts on the one of the most animated online opinion forum “PTT” BBS of Taiwan. The research use the traditional Bag-of-Words model, Latent Semantic Analysis and Word Embedding methods, coupled with Machine learning Support Vector Machine and Deep Learning Long-Short-Term Memory network and Convolutional Neural Network topology structure to establish the mode of stock price prediction model. The experiment conducted the model training and testing with the label of the stock price changes on the next trading day after the news release and the subsequent trading days. The results of the experiment found no obviously, directly correlation between the both variables, but still can track the fluctuations of the stock price through tiny signals. Due to the insufficient amount of raw data, the model inevitably appears to be over-fitting for training datasets. In the future research, suggested that the model design combining quantitative data, the exploration of social media messages, the definition of Y variables, and the inclusion of sentiment analysis.