Stock Prediction Research Based On Investor Sentiment

As an important part of the financial market, stock has played a very big effect in the process of economic development. Since 2015, with the fluctuation of Chinese A-share market, investors’ sentiment also begins with a passion to a more rational trading. Therefore, it is becoming more and more popular research topic to study stock forecasting, and give reasonable investment advice. Traditional stock prediction research is usually based on the relevant technical indicators of the stock market, And the impact of noise factors,such as investor sentiment, is usually reflected indirectly by the indicators of the stock market,which has usually not particularly good results for stock forecasting.With the development of network technology within the core Web 2.0, more and more investors begin to publish their opinion of the stock market and exchange opinions with others on the web, these original opinions provides an opportunity for our studyon investor sentiment. We downloaded stock comments from stock forum firstly and divided these comments into positive, negative and neuter by support vector machine(SVM), then we calculated sentiment index and divergence index of each day for every stock by statistical methods. Finally, we constructed the MI-SVM prediction model and SS-SVM prediction model based on these two indices, and predicted the market closing index and part of the stock’s closing price using support vector machine regression analysis, it is concluded that the SS-SVM model is superior to the MI-SVM model.The main research work of this thesis can be summarized as follows:(1)Based on the basic principle of sentiment classification and the theory of vector space model, we proposed a feature weighting method that combines the structure of the document, this method is used in sentiment classification and better results are achieved.(2)With the attention indicators constructed by posts’ page views and number of replies and the emotion tendency measured by sentiment classification results, we calculated sentiment index and divergence index comprehensively.(3)After understand the principle of stock prediction, two stock index prediction model is constructed, which MI-SVM considered the basic technical indexes of the stock market, and SS-SVM considered investor sentiment additionally, then we compared the fitting degree between the two prediction models’ results and the real values. In this paper, we get the conclusion that investor sentiment index is beneficial to stock prediction.