Research On Corporate Financial Early Warning From The Perspective Of Annual Report Text Mining

Enterprise bankruptcy is often regarded as the definition of enterprise failure.When the financial distress occurs,the company cannot repay debts on time,and shareholders and creditors suffer significant property losses.However,the occurrence of corporate financial crisis is not sudden,but is usually caused by slow accumulation and final outbreak in daily operation,which makes corporate financial distress has the characteristics of predictability.The purpose of financial early warning is to send out warning signals before the financial distress of the company by analyzing the obtained information.The annual report issued by listed companies is an important source for external stakeholders to obtain corporate information,and the text information in the annual report has not been concerned by related users for a long time.In this study,the listed companies in China’s A-share Shanghai Stock Exchange and Shenzhen Stock Exchange from 2015 to 2019 are selected as the research samples to explore the corporate financial warning model from the perspective of annual report text mining.The text mining of annual reports measured by management text tone and annual report text tone as the starting point,studies the financial early-warning ability of annual reports under different industry categories,and the moderating effect of traditional financial distress early-warning indicators and audit opinion variables on the financial early-warning ability of annual reports.In this paper,literature review research method,normative research method and empirical analysis method were used to carry out the subject research,in which the empirical research respectively adopted basic regression analysis,grouping regression analysis,moderating regression analysis,replacement model regression analysis and other methods.The research conclusions were drawn:In basic regression analysis,the annual report text mining under the introduction of the annual report text tone will improve financial early-warning model fitting degree and the ability to predict.However,as the complexity of economic activities and characteristics of the disclosure text,there is a time gap between the report releasing and the occurrence of financial distress.In the further regression analysis,it is found that the financial distress warning ability of the annual report tone and the tone of management text is various between the industries.The manufacturing industry,information service industry and real estate industry have the strongest financial distress warning ability.There was a negative correlation between annual report tone,as well as the management tone,and the financial warning.This study further found that the annual report tone and the tone of the management are negatively correlated with the occurrence of financial distress,that is,the more positive the tone of the annual report and the tone of the management are,the worse the enterprise’s operating condition is likely to be,and.the more likely the enterprise is to have the risk of financial distress.The auditor’s opinion will restrain the over-optimism of the annual report text tone and the management text tone,and further revise the financial warning model.The study in this paper is still robust after model replacement.In this paper,through in-depth mining of the text information of the annual report,it has a certain enlightenment function for the users of the annual report information how to use the information of the annual report more efficiently and carry out financial distress warning from multiple perspectives.This paper makes a comparative analysis of the annual report tone and the tone of management in different industries,and further revises the financial early warning model from the perspective of external audit opinions.At the same time,it also shows new research ideas for relevant scholars engaged in the study of text analysis.