An Empirical Analysis On Stock Price And Investor Sentiment Based On Stock Community Text

With the popularization of investment knowledge and the development of the Internet,more and more individual investors have entered the stock market through the Internet.At the same time,the change and prediction of stock prices have become a hot topic among people on social networking sites.As investors continue to pile in,overall investor sentiment also plays a role in driving stock prices.Therefore,this paper took this phenomenon as the entry point,collects the text from the Internet stock community,builds a stock price prediction model based on the investor sentiment information mined from the text and the historical trading information of some certain stocks.In the analysis of this paper,we first crawled tens of thousands of stock comment data from the stock community based on the Scrapy framework,and stored them in the database after filtering and cleaning.The parameters of several models were tuned by manual annotated data,and the Naive Bayes was finally determined as the sentiment polarity classification model.A sentiment classifier was used to carry out text sentiment polarity classification on the crawled data.After the classification was completed,the sentiment polarity was used to construct the investor sentiment index,and the time nodes were re-divided.The calculation method of dividing the natural day was changed to 9 O'clock per day,and the sentiment index in the holiday period was combined to calculate the next day.The index showed a significant weak correlation with the closing price.In order to alleviate the problem of vanishing gradient in the long time interval series,the Long Short Term Memory network was used to build the price prediction model.The model was trained by historical trading information and the investor sentiment indicator,and the optimal parameters were determined.Based on this model,the stock price variation tendency could be predicted,and the prediction accuracy of the direction of rise or fall in the next trading day was close to 60%,which was significantly improved compared with the benchmark.