Correlation Analysis Between Investor Sentiment And Stock Price Trend Based On Text Mining

As the network gets a rapidly popularity in people’s work and life, the Internet has become a variety of information collection and distribution center. The network information not only changes the way of investment for investors, but also becomes the basis of information exchanges and investment decision-making for investors. So accessing to information in the online forum is an important way of understanding investor’s psychology and behavior, which has practical significance on the stock market. We make the sentiment analysis on stock reviews in BBS on the basis of behavioral finance theory and limited attention theory, and then study the influencing relationship between stock price changes and investors’ sentiment.Firstly, we make use of text classification to extract relevant stock reviews, and then make use of sentiment classification to analyze investor’s emotional tendency(bullish and bearish). Irrelevant information are identified in the process of eliminating irrelevant reviews, which not only ensured irrelevant reviews are basically absorbed, but also ensured the accuracy of data analysis on stock reviews. In the subsequent process of sentiment analysis, we compares the semantic analysis methods and support vector machine, and support vector machine combined with the information gain method can obtain better experimental results than the semantic analysis methods.Secondly, we select proxy index which can stand for investors’ sentiment, make use of principal component analysis to characterize comprehensive sentiment index, and use granger causality test and multiple regression models to study how investor sentiment influence the stock yield rate. The results showed that investor sentiment is one of factors that influence stock yield rate, and have positive influence. In addition, the current investor sentiment and fluctuant investor sentiment have more influence on stock yield rate than previous investor sentiment.