Research On The Performance Forecast Of The Company Based On The Discussion And Analysis Of Emotional Intonation By The Management Of Annual Report

The data produced in the financial field are growing rapidly at an amazing speed,and a large and complex variety of data are accumulated.However,the "rich data can be valuable and poor" has become a common phenomenon.How to find valuable decision information from the massive financial field data is a problem that many financial analysts and market investors are interested in.Annual report is used to describe the whole operation of the company in the current year.It includes both text content and basic information of the company,distribution of shareholders,important matters and other aspects,as well as the financial statement data of the company.The emotional tone of management discussion and analysis in annual report is not only the focus of annual report processing,but also the basis for analyzing the performance of the company.The combination of annual report management discussion and analysis of emotional tone and annual report financial data can help investors to quickly understand the company's operation and make efficient and accurate decisions.This paper mainly focuses on the problems existing in the traditional annual report analysis to verify the effectiveness of the management discussion and analysis of emotional intonation and the accuracy of performance prediction of the company.The main research contents are as follows:(1)In the traditional annual report,the sample data is limited,usually the annual reports of a certain industry or several companies in recent five years.The data set is small and incomplete,and the research results are biased.This paper uses crawler technology to expand the research data set,draws on the previous researchers' literature and takes into account the problems of two kinds of positive and negative emotion dictionaries,and selects five kinds of annual report emotion dictionaries,which are positive,negative,blind positive,uncertain and deterministic emotion dictionaries.By using text mining technology,we get two kinds of emotional intonation TF and TF-IDF,which are discussed and analyzed by the management of the annual report.Through modeling with naive Bayes model,we find that adding the emotional intonation feature data of the management of the annual report to the financial data of the annual report can improve the relevant indicators of the performance prediction model to a certain extent,Especially based on the naive Bayes model,when the financial data of the annual report remain unchanged,adding the emotional tone of discussion and analysis of the management of the annual report to predict the company's performance,the accuracy rate has increased from 53.42% to 61.27%,which has increased by nearly 8%,proving that the emotional tone of discussion and analysis of the management of the annual report has improved the efficiency of the company's performance prediction model.(2)The traditional research on emotional tone and performance prediction of annual report management is based on the correlation analysis.It does not forecast and analyze the company performance by combining the discussion and analysis emotion tone of annual report management with annual report financial data.This paper classifies the company performance by comparing the company performance horizontally.The paper uses the support vector machine and other related models to form a new annual report management discussion and analysis emotion tone based on the combination of annual report financial data and annual report management discussion and analysis emotional tone to predict the company performance,which can further analyze the development of the company performance.(3)In order to further improve the generalization and accuracy of the whole forecast company performance model,this paper uses the integrated learning idea to build a more efficient and better performance integrated learning and forecasting company performance model,and then the predicted values from the training of multiple forecast companies' performance models are re integrated.Compared with the optimal basic model,ACC and AUC of the integrated learning model have improved by 2 percentage points,thus improving the performance of the whole company performance prediction model.