Research On The Influence Of Network Investor Sentiment On Stock Market Return

Investor sentiment refers to the systematic deviation of investors' psychological expectation of the future market,which is likely to lead to irrational investment behavior.If investor's bad sentiment spreads through the crowd,it could have a serious impact on the market.It is of great significance to accurately measure investor sentiment and study its influence on the development of the stock market that it also has gained the attention and research of the academic circle and the industry.However,the traditional measurement methods,such as questionnaire survey,are time-consuming and laborious with large errors.With the development of social networks,individual investors tend to view and exchange stock market information on the Internet,making the Internet a resource for investor sentiment.Therefore,in this paper,I build the multiple sentiment indexes from the perspective of investor attention,emotional change,and the difference degree of emotional change.It's all built on the emotion analyzing of text data from network.Those indexes measure the investor sentiment for the network comprehensively.Then I analyses its degree on the yield rate and fluctuation of yield rate.In the end I find that the network investor sentiment can improve the prediction effect and warning effect.The specific research work and conclusions are as follows.One is the measurement of network investor sentiment.This paper first grabs the text data of Donfang fortune online stock exchange index bar from August 20,2015 to May 24,2019/5/24,then perform data filtering and cleaning using Rwordseg of R language and customizes word segmentation thesaurus for word segmentation and feature extraction.Then perform text emotion analysis based on the custom stock domain emotion dictionary.Finally,this paper constructs indexes not only from the perspective of investor attention,but also from the changes in investor sentiment.These indexes are the overall attention,the difference index of positive and negative emotional attention,the positive expectation index,the negative expectation index and the emotion expectation dispersion index.This paper uses the five indexes constructed to measure network investor's sentiment in a multidimensional and comprehensive way.The second is the research on the influence of network investor sentiment on stock market returns.Firstly,through correlation test,it is found that difference index of positive and negative and the overall attention is significantly positively correlated with the stock market turnover.Then,this paper uses a series of methods and the ARMA-GARCH-M model to find that network investor sentiment significantly affects stock market returns and earnings volatility.Among them,the positive expectation index has a significant positive effect on stock market return,while the negative expectation index has a significant negative effect on stock market return,and the influence degree is not symmetrical.The influence degree of negative emotion is 35 times higher than that of positive emotion.Thirdly,the research on the influence of network investor sentiment on stock market return prediction and early warning effect.In the first place,support vector machine(SVM)is used to build the prediction model of stock price.Traditional stock market performance indicators and the network investor sentiment have been introduced successively and the comparison showed that the prediction accuracy is improved by 9.2%.Again this paper defines the risk of stock market,Secondly,according to the standard that the sample with the lowest rate of return accounts for about 10% of the total sample as the risk sample,the binary variable method is adopted to define the stock market risk.After this this paper uses synthetic minority class sampling(SMOTE)method to eliminate the influence of the unbalanced sample data of SVM classification.At length;this paper builds the stock market risk early warning model introducing the traditional stock performance indicators,the prediction accuracy is recorded.It is found that warning accuracy is improved by 11.2% when the network investor sentiment is introduced.It is indicated that the network investor sentiment can significantly improve the stock price forecast and risk warning effect,which is helpful to provide a reference basis for rational investment of equity investors and monitoring stock market risk by regulatory authorities.