Study of Stock Prediction Based on Social Network

The study on the interactions between social media and financial markets is an interesting topic. This paper is to investigate this issue for stocks from the Shanghai/Shenzhen stock exchange, based on a popular online Chinese stock forum Guba.com.cn. Other than the traditional sentimental analysis method, for each stock, we build a Social Behavior Graph based on human´s online behavior, calculate some key characteristics of the graph, and find out the correlations between trading volume/price and those characteristics. Furthermore, we make use of a BP-neural network to predict the trading volume and price of stocks. Our method has achieved a better outcomes compared to the traditional trading volume/price based time series models. A trading strategy based on our method achieved 56.28% benefits in only three month´s time, when the stock price just increased 1.17%.