Research On Personal Credit Risk Assessment Model Based On Machine Learning

With the continuous growth of the consumer finance and the diversification of business development types recent years,how to quickly and accurately fully evaluate customer credit risk has become the core competitiveness of consumer finance companies.Machine learning has been widely used in consumer finance recent years.It is of great practical significance to study the application of machine learning in consumer finance credit risk assessment.This paper studies the current situation of credit system construction,empirically analyzes the data of an Internet bank.With different data sources in the personal credit system,trained in Logistic Regression,Decision Tree,Random Forests,XGBoosting,and LightGBM model,to build the submodel.Then the submodel outputs scores as input variables,to built a stacking model.All models assessed with various standard to compare the advantages and disadvantages of these algorithms theory and business rationality aspects.Through this research,it can be concluded that the Logistic Regression model’s identification ability of credit default risk is obviously weaker than that of machine learning.Ensemble predictors is stronger then a single predictor,and stacking model is much more better then ensemble models.This paper establishes a stacking model to verify that machine learning algorithm is the core technology of consumer credit risk assessment model,which has a certain business and technical reference significance for the study of personal credit risk scorecards.