The Credit Default Prediction Model Of Companies Based On The Optimal Ensemble Of CART

Credit risk is default risk,refers to the possibility that the borrower is unwilling or unable to return loans,causing losses.The credit default prediction model of companies is to predict the default probability of the company in the future.According to the results of default prediction,the company's stakeholders can adjust their investment strategies to avoid credit risk and credit crisis.This study consists of five parts: The first chapter is the introduction.The second chapter is the basic principle of the default prediction model based on the optimal ensemble of CART.The third chapter is the construction of default prediction model based on the optimal ensemble of CART.The fourth chapter is an empirical study of Chinese listed companies.The fifth chapter is the conclusion.The research focus of this study includes: Firstly,in a typical random undersampling method,if the ratio of the number of defaulted and non-defaulted samples is different,the prediction accuracy of the model is different.Objectively,there must be an optimal sample ratio to ensure that the model has the highest accuracy in default prediction.Secondly,in the ensemble of CART model,if the number of decision trees is different,the prediction accuracy is different.Objectively,there must be an optimal number of decision trees,so that the prediction accuracy is the highest.The third is the prediction period of the model.For example,the loan decision occurred in the past or the present,and the repayment of the enterprise occurred in the future.It seems that there is no point in using the current year's data and the default status of the enterprise in the current year.The model's default prediction period is too short and has little effect.The characteristics and innovations of this research include: Firstly,in the case of traversing the proportion of two types of samples between defaulted companies and nondefaulted companies,we reverse the optimal ratio according to the prediction accuracy of the model.In ensemble of CART model,we reverse the optimal number of CART according to the prediction accuracy of the model.And we set the credit default prediction model of companies based on the optimal ensemble of CART.Secondly,based on the existing research forecast period of 3 years,we model by using the t-k year indicator data and the t-year default status.We set a model that predicts the default probability of the t+k(k=1,2,3,4,5)year using the data of the t year,and expands the forecast period of the model.By comparing with typical predictive models,our model has the highest accuracy and the longest forecast period.Our study found that the credit characteristics of Chinese listed companies are: The industries of “transportation,warehousing and postal services” and “information transmission,software and information technology services” have the best credit qualifications.The industries of “mining industry” and “real estate industry” have the worst credit qualifications.First-tier cities such as Beijing,Shanghai,and Shenzhen,have the best credit qualifications.Four-tier city companies such as Zhoushan,Kaifeng and Baoji,has the worst credit qualifications.