Research On The Impact Of Investor Sentiment On Stock Market Returns Based On BERT Model

With the development of the Internet,online platforms have expanded the reach and ability of investor sentiment.Nowadays,investors not only express their own views on market conditions on online platforms,but also make decisions based on the information on the forums,which makes investors inevitably influenced by the sentiments of other investors on the Internet.However,the management of online platforms is not standardized,the information on the platforms is mixed,and false information is frequently published.Strengthening the supervision of online public opinion and guiding investors to view online information rationally will help maintain the stability of my country’s financial market.This paper uses the stock review text of China’s largest investor forum,Oriental Fortune Network,as the data source of investor sentiment,and uses the BERT model as a measurement tool for investor sentiment to construct an investor sentiment index.In-depth research on the relationship between market returns broadens the technical means of investor sentiment measurement,and provides an empirical basis for the relationship between investor sentiment and stock market returns based on web texts.After summarizing the relevant literature on investor sentiment,this paper sorts out the development process of the measurement method of investor sentiment,and introduces the transmission mechanism of the impact of investor sentiment on stock market returns.Scholars have not yet reached an agreement on how to measure investor sentiment,and the measurement of investor sentiment is still a key premise for investor sentiment research.This paper adopts the most advanced BERT model in the field of natural language processing(NLP),and conducts Internet stock reviews on the BERT model.Targeted pre-training for sentiment classification tasks,including stock rating dictionary coverage,feature information enhancement,and addition of LSTM units.By comparing the classification results,it is found that the classification performance of the adjusted pre-trained BERT model has been significantly improved.The data constructs the investor sentiment index,and the empirical results show that the investor sentiment index constructed in this paper has a higher degree of fit with the stock market returns and better prediction ability.In order to verify the mutual influence between investor sentiment and stock market returns,a vector autoregression model was constructed based on investor sentiment indicators and stock market returns.The empirical results show that investor sentiment and stock market return have a positive impact on each other.The positive impact of stock market return on investor sentiment lasts for a short period of time,and investor sentiment may reverse in the long run,leading to a disposition effect.The investor sentiment has a long-term positive impact on stock market returns,and the mutual influence of the two has a magnifying effect.The conclusion of this paper provides an empirical basis for explaining market anomalies such as amplification effect and disposition effect,and demonstrates the mutual influence between investor sentiment and stock market returns,which is of great importance for the study of the relationship between investor sentiment and stock market.theoretical significance.This paper provides a new technical means for the measurement of investor sentiment,and discusses the evaluation standard of the investor sentiment measurement method.The research on investor sentiment measurement has practical significance and application value.