The Research And Implementation Of Bank’s Overdue Forecast Model For Public Client Loans

At present,China’s financial market is seriously unstable.As the main financial intermediary,the bank’s sustainable,healthy and stable development is closely related to the stable development of the country.Public credit business as the main business of the bank has always brought high profits to the banking industry,but behind the huge amount of loans,there is also a huge risk of default.Traditional customer default risk assessment only uses basic customer information and loan-related information.However,there is a complex social relationship between customers and customers.In this thesis,we find that the guarantee factor has a greater impact on customer loan overdue through statistical analysis.Therefore,based on the enhanced learning model XGBoost,we construct an XGBoost model driven by guarantee relationship(XGBoost-GUAR).The main work of this thesis is as follows:(1)Establish the index system of loan overdue risk assessment,collect the data of loan overdue risk assessment,and carry out data preprocessing and feature processing.(2)Establish XGBoost model with pre-processed data,construct customer guarantee network and discover potential features by community detection,then add guarantee factors to modify the model results,and finally establish guarantee relationship-driven XGBoost model(XGBoost-GUAR).(3)Complete the visualization of customer loan overdue forecast results,visualize overdue customer lists and customer guarantee network overdue risks.In this thesis,the XGBoost-GUAR model and the traditional prediction model are compared.The results show that the model has better accuracy.The customer loan overdue prediction system based on this model has been applied in a commercial bank and achieved good results.