| With the continuous development of the personal credit industry,the complex financial scenarios and customer group structure make the Internet loan platf orm gradually face more problems that are difficult to solve in the actual lending business.The impact of changes in the macro environment on customer default risk and the uncontrollable change of personal repayment ability over time after advanced consumption have become important considerations in the study of credit default prediction.At present,domestic and foreign scholars’ research on credit default prediction not only focuses on the basic information of customers,but also ignores the impact of macroeconomic and time-varying characteristics on customer default risk,which has certain limitations and lags behind.Therefore,based on the credit account data set provided by the real business of PPDAI,from the perspective of adding effective factors,and considering the three dimensions of customer’s personal account characteristics,macro characteristics and time-varying characteristics,four default prediction models are constructed by using the research method of dynamic survival analysis,and the i mpact of the added dimensions on customer default risk is judged through comparative analysis,And further,according to the hazard ratio,we can judge whether the influencing variables are risk factors or protective factors,in order to construct a forwar d-looking prediction model that comprehensively considers the macroeconomic and time-varying characteristics.The empirical results show that GDP,CPI and disposable income of urban residents are protective factors,PPI is a risk factor,and the impact of loan term and age on the borrower’s default risk changes over time.Based on the research results,we can see that the macro characteristics and characteristic time dependence do affect the customer default risk,and the impact of time-varying characteristics on the default risk will change over time.Adding macro factors to the loan default prediction model and considering the time-varying characteristics can enable the Internet loan platform to comprehensively predict the borrower’s default probability at both macro and micro levels before the loan,avoid the decline in default prediction accuracy caused by ignoring time-dependent variables,and monitor the default risk fluctuation in real time after the loan;It is also helpful for the regulatory authorities to grasp the expected overdue loan situation in the industry,control the degree of risk deterioration,and avoid industry impact. |
