Research And Design Of Credit Score Card Based On Machine Learning

Consumer finance is in a good state of development,especially the rapid development of Internet financial platforms relying on traffic advantages,providing consumer financial services for a large number of customers.However,in terms of risk control,compared with traditional credit risk management,Internet consumer finance is not limited to the data of traditional financial ability,but also includes massive data of customers’ social behavior,consumption habits and other multidimensional characteristics.The efficiency of the traditional risk control model is challenged and may not be able to adapt to the credit risk management of Internet consumer finance under the background of big data.In order to effectively solve the problem of credit risk management of Internet financial platforms,this paper designs a credit scoring product suitable for credit risk management of Internet consumer finance based on machine learning algorithm and data of Lending Club,the largest P2 P online lending company in the United States.Firstly,through theoretical analysis and feature processing,this paper selects the factors that may have an important impact on customer default.Secondly,the traditional Logistic regression and tree model-based integration algorithms XGBoost and Cat Boost are used to build a personal credit risk default prediction model.Then,according to the AUC model effect index,the Cat Boost model is used to design credit rating products,and the interpretability of the model is innovatively discussed.How each feature affects the prediction result of the model is analyzed.Finally,the standard of credit risk scoring card is generated,and then the distribution of customer scores is associated with the rating to generate the final credit scoring product,and the implementation and application of the scoring card is analyzed in each stage.This paper attempts to further optimize the scoring model and introduce advanced machine learning algorithms to construct and design a set of interpretable Internet consumer finance credit scoring products suitable for the context of big data.This credit scoring product has achieved good prediction effect and realized the matching of platform risk and income,which has certain reference significance for other Internet consumer financial institutions and platforms to improve the automation and efficiency of personal credit risk management.