Default Prediction Model Of Listed Company Based On Business Plan Text

Default prediction of listed companies refers to the use of structured and unstructured data of listed companies in the past to predict the default status of listed companies in the future.Accurate prediction of default status of listed companies provides important support for investors,banks and credit institutions of listed companies,regulators and other decisions.This paper constructs a default prediction model based on business plan text,including four chapters: the first chapter is the introduction,the second chapter is the construction of default prediction model,the third chapter is the empirical analysis,and the fourth chapter is the conclusion.The research focus of this paper includes two aspects: one is the choice of variables.Variable is the basis of default prediction model construction.If variables containing different information are input into the same model,the default prediction results may be quite different or even completely opposite.The second is the assignment of intonation variables.Different assignment methods,variable size and impact on default prediction are different,so there is bound to be a better assignment method to distinguish defaulting and non-defaulting companies.The innovation and characteristics of this paper are as follows: first,in the selection of predictive variables,on the basis of numerical variables,we add overdue and violation extracted from the operation plan text of the annual report of listed companies.The prediction model of text variables in the form of accident is established to reveal the joint influence of text variables and numerical variables on the default status,which improves the prediction accuracy of enterprise default.Two is on intonation variable values determine,through different kinds of emotion in the logistic regression equation of the influence of the number of words to the default size to weighted number of emotional vocabulary,tone variables to determine the value,reflect the tone variables and the default of direct contact change the existing studies intonation variable assignment has nothing to do with the default phenomenon.The data of Chinese listed companies from 2 012 to 2 020 are studied.First,with the addition of text variables,the prediction accuracy based on AUC increases by 13.1%-19.0%in six prediction windows m=0,1,2,3,4,5.Second,there are 8 variables with 0-5 years’ predictive ability for default of Chinese listed companies,including "asset-liability ratio","liquidity ratio" and "positive intonation variable POS".There are 15 variables with 0-2 years’ short-term predictive ability,including "negative intonation variable NEG" and "net intonation variable TONE".There are 12 variables with 3-5 years of long-term forecasting ability,including three text variables,"brand","profit" and "positive intonation variable".