| Public-Private Partnership(PPP)is a long-term collaborative mechanism in which the government and social capital jointly provide infrastructure construction or public services.This mechanism is based on a concession,and the rights and responsibilities of both parties are contractually agreed upon,forming a risk-sharing and profit-sharing partnership.Currently,the risk of debt default by PPP project companies has become a potential threat to the financial system.Effectively preventing this hidden risk is one of the key issues in PPP financing research.In recent years,the use of machine learning in corporate credit evaluation has become increasingly popular.However,most studies only focus on the output results of the model,without explaining in detail the degree of variable influence and decision-making process.Therefore,identifying the key factors of default risk in urban investment enterprises,constructing corresponding credit evaluation models,and scientifically explaining the evaluation results have become critical issues in this research field.Establishing a corresponding credit evaluation model is to predict the possibility of default risk,enabling PPP market entities to make risk decisions that minimize losses in response to credit crisis signals.In view of the high research interest in debt default risk related to PPP projects and poor solvability of traditional corporate credit evaluation models,this paper conducts data feature-driven research on intelligent credit evaluation models for PPP project enterprises.Firstly,financial data of urban investment enterprises were analyzed for correlation analysis to understand their characteristics of default risk and more accurately construct feature engineering.On this basis,an XGBoost-Logistic combination algorithm was established by integrating typical features of PPP projects and taking advantage of comprehensive gradient boosting algorithm model classification performance and high solvability logic regression model.An intelligent credit evaluation model for PPP project enterprises was constructed which can accurately identify key factors affecting default risks and give a credit score.This paper also uses SHAP value analysis to explain the key feature indicators that affect the results of the credit score,and applies Dtreeviz model visualization technology to visualize the decision-making process and calculation method of overall enterprise credit evaluation models.To verify the feasibility and accuracy of the model,K-fold cross-validation method was used along with gain boosting performance evaluation method.The verification results show that compared with single evaluation models,our constructed enterprise credit rating model has better stability and estimation accuracy;XGBoost-Logistic algorithm can achieve optimal classification effect covering about 70% samples.This indicates that our proposed model can comprehensively evaluate the credibility status quo for PPP project entities systematically while improving its technical level in terms of credibility assessment techniques.Finally,based on our proposed intelligent enterprise rating system using Spring Boot framework as well as other technologies such as machine learning algorithms etc.,we developed an intelligent platform for evaluating PPP project companies’ credits which provides effective ways for identifying urban investment company’s default risks thereby assisting banks and government departments in providing more accurate loan support policies and measures.This not only helps to improve the accuracy and predictive ability of PPP project enterprise credit evaluation,but also provides important basis for banks and governments to formulate more targeted credit policies and risk prevention measures.Therefore,establishing a credit rating model has extremely important practical significance and application value for both PPP market entities’ credit risk control as well as bank or government supervision work on PPP project enterprises’ credits. |
PDF Full Text Request