Research On The Influence Of Investor Sentiment On Stock Yield Of GEM Market

China’s GEM market was officially listed in October 2009 and has been developing for more than 10 years.From gem since the opening,with a rapid development of the attitude forward.However,there are still many unfavorable factors to its long-term development,such as "excessive trading" and "herd effect".In addition,the enterprises in gem are mostly emerging small and medium-sized enterprises and high-tech enterprises.Due to its characteristics of high growth and high risk,the stock market has a high price/earnings ratio,and the future forecast trend is also uncertain.With the rapid development of Internet media,investors express their views and spread their emotions through stock bar and other new media.Investor sentiment is the transfer station of information in the stock market and financial commentary,presenting the emotional state of investors,and a large number of mixed information will cause investor sentiment fluctuations.Therefore,the in-depth study of whether such information value has an impact on the stock return rate of gem in China has certain reference value for the stable development of the stock market,and to a certain extent enriched the research in the field of behavioral finance in China.Therefore,based on the perspective of text data analysis,this thesis constructs investor sentiment index to test how it affects the stock return rate of gem in China.The main research contents include :(1)The concepts related to investor sentiment,stock return rate and text data mining are expounded,and the research literatures of domestic and foreign scholars are introduced.(2)Put forward the research hypothesis based on the analysis of relevant theoretical basis.(3)Use Python to crawl the comment data of the stock bar,analyze the comment content by using LSTM sentiment classification model,and then build an investor sentiment index suitable for research.(4)At the macro level,the constructed investor sentiment index is combined with the return rate of GEM index,and the VAR model is used to test and analyze the causal relationship.(5)At the micro level of listed companies,through the fixed effect model,the correlation analysis of the constructed macro level investor sentiment and the stock return rate of individual stocks is made.(6)Based on the research content,this thesis puts forward some referential suggestions for investors in the stock market,listed companies and relevant government regulatory departments.Through the combination of theoretical analysis and empirical test,the following conclusions are drawn:(1)Through the analysis of the text data to build investor sentiment index,to some extent,reflects the overall trend of the gem market investor sentiment in our country,and from the macroscopic and microscopic two levels the empirical test shows that the index on to explore the influence on the growth enterprise market stock yields have a certain rationality and research.(2)The VAR model test results show that the investor sentiment index constructed based on text data analysis and the RETURN rate of GEM index are granger causes for each other.For the impact of the return rate of GEM index,the biggest factor is itself.Even though the impact of investor sentiment is relatively small,it cannot be ignored.(3)At the micro level,the use of fixed effects model,the empirical results show that investor sentiment has obvious positive effect on individual stocks yields,and the index of affiliated companies in 14 industries effects of individual stocks yields vary,manufacturing,high and new technology industry and agriculture,forestry,animal husbandry and fishery company stocks yields more significantly than other industries affected by investor sentiment;Compared with non-state-owned companies,state-owned companies are more susceptible to investor sentiment.Large companies are more affected by investor sentiment than small,medium and micro companies.Based on the research conclusions of this thesis,the content of behavioral finance can be enriched and applied locally to some extent,and the work of big data management and information dissemination can be warned to some extent.