A Study Of Listed Company Periodic Report Based On Text Sentiment Analysis

The periodic report of listed companies is one of the main forms required by China Securities Regulatory Commission to disclose information to the society and public.As an official data source for the society and public to obtain information related to listed companies,periodic reports of listed companies are a major information channel for discovering investment opportunities and can provide reference information for investment decisions.With the increasingly strict supervision of China's securities market,the content of listed companies' periodic reports is becoming more and more authentic and the format is becoming more and more standardized.The section of "Management Discussion and Analysis" in the periodic report of listed companies explains the next year's business plan and the opportunities,challenges and risks faced by the company's future development from the perspective of the company's management,which also attracts more attention from investors.Therefore,the text mining of listed companies' periodic reports is an important means for investors to deeply understand the company's operating situation and accurately grasp the company's future development direction.In this paper,data mining technology is used to process the mid-term reports and annual reports issued by about 3,000 listed companies on the main boards of Shanghai and Shenzhen Stock Exchange from 2016 to 2020,totaling more than 20,000.The research objects include listed companies' financial indicators and "Management Discussion and Analysis".The combination of quantitative data and qualitative data can well reflect the information disclosed in the regular reports of listed companies.This paper aims to conduct text sentiment analysis on the "Management Discussion and Analysis",to discuss the relationship between the future business situation of listed companies and the management emotion information,and on this basis,to establish a model to predict the stock trend of listed companies.The research content of this paper is mainly based on the application of text sentiment analysis in the periodic reports of listed companies.Firstly,the emotional dictionary is constructed to quantify the emotional information contained in "Management Discussion and Analysis".Then,through regression analysis,it is confirmed that the positive and negative tendencies of management emotional information will have different degrees of impact on the operation of listed companies.Finally,the machine learning model is used to accurately predict the stock trend of listed companies.According to the research conclusions drawn from the empirical analysis in this paper,investors can make more scientific investment decisions while finding more investment opportunities;Listed companies can disclose information more accurately and efficiently,and improve the quality of information disclosure;Relevant regulators can formulate more targeted rules and regulations on information disclosure and optimize the information channel between listed companies and investors.