Research on the Prediction Model of Borrowers’ Severe Delinquency Behavior Based on Multi-Dimensional Credit Indicators

In the context of the expansion of the consumer credit market, borrowers’ credit behaviors have become diversified and complicated. This study is committed to integrating multi-dimensional credit indicators such as revolving credit utilization rate, debt-to-income ratio, and overdue records of different periods to build a high-precision prediction model, so as to evaluate the probability of borrowers’ severe delinquency behavior in the next 2 years. The research uses 16,715 personal loan records from openML with dataset ID 46543 as data support, applies a linear regression model for analysis, and carries out the research by defining variables such as credit risk, debt repayment risk, and credit activity-to-age ratio. The results show that variables such as credit activity-to-age ratio, real estate debt dependence, and severe delinquency prediction have a significant linear relationship with the dependent variable, and there is no multicollinearity among the variables. Finally, the corresponding regression equation is obtained. This study provides a scientific basis for financial institutions’ risk management, credit approval, and post-loan monitoring, and helps to reduce the non-performing loan ratio and optimize the allocation of credit resources.