Research On Personal Credit Risk Assessment Based On Machine Learning Method

After nearly a decade of explosive development,my country’s consumer credit market has gradually transitioned to a new stage of steady growth.In the context of dual circulation,the importance of consumer finance remains unchanged.At the same time,the causes of credit risk have become more complex and have a greater impact on the overall economy.Traditional credit evaluation methods have problems such as difficulty in handling large-scale data and insufficient evaluation accuracy.Therefore,credit risk evaluation has gradually shifted from qualitative research to statistics,machine learning and other methods.How to construct an efficient and accurate personal credit evaluation method has become a current hot research field.Existing machine learning techniques,such as decision trees,support vector machines,etc.,have shown better performance than statistical models on credit risk assessment problems.Big data provides a new idea for personal credit risk assessment in consumer finance.Integrate and analyze the multi-dimensional behavior data of users on the Internet to build a richer user portrait.Relying on social big data,the research on personal credit risk assessment based on machine learning method has important theoretical and practical significance.Based on real Internet consumer finance and consumer credit data,this thesis incorporates new Internet multi-dimensional data,initially establishes a credit evaluation model based on three methods of logistic regression,GBDT,and LightGBM,and uses precision rate,recall rate,AUC for comparative analysis.Experiments show that the risk assessment model based on LightGBM method has certain feasibility and research value in practical application scenarios.Under the premise of ensuring precision and certain recall rate,it can effectively identify personal credit risk in consumer credit,and recall risky users.