A Study On The Internet-Forum-Based Investor Sentiment

As current society and economy progress continually, more and more individual investors enter into the financial market and constitute a non-negligible group. Meanwhile, as the combination of financial investment and the Internet is getting closer, an increasing number of investors tend to express their emotions on internet platforms. Therefore, it is particularly necessary to study on investor sentiment in the Internet context. Then, what are the behavioral characteristics of individual investors on the Internet? What is the logical relationship between these characteristics and dissemination of emotional information? How to learn about investors'emotional tendencies on the Internet and build reasonable investor sentiment indexes? These are the problems that interest participants of the financial market and this study hopes to solve.In previous studies, the images of China's individual investors are relatively fuzzy. After an in-depth analysis of related theories, this paper issued over 20,000 online questionnaires in the form of questionnaire to know about behavioral characteristics of China's individual investors and clarify the relationship between investors' characteristics and dissemination of emotional information. In addition, in this study, how to obtain investors' sentiments is also a difficulty. So, this paper utilized data extracting tools to acquire over 1,180,000 forum posts that span two years and other additional information, and extracted text messages in these posts with emotion recognition techniques. Finally, this paper proposes a 3D-vector model for measuring investor sentiments by combining theories of psychology, and constructs investor sentiment indexes with approaches of neural network machine learning.Through this study, we acquired a preliminary overview of behavioral characteristics of individual investors in China's mainstream financial internet forums, and successfully extracted and utilized emotions of online investors. It is evidenced that internet-forum-based investor sentiment indexes can better reflect real financial markets, and adding investor sentiment indexes can significantly improve the precision of predicting rate of return of Shanghai Stock Exchange (SSE) by the neural network model. Also, investor sentiment indexes constructed by this study also have predictive validity about closing price of SSE and of certain practical value for the financial industry.