The Study On Credit Rating Of Personal Credit Loan In Bank A Based On Machine Learning

Bank A is a traditional commercial bank in Sichuan Province.For many years,its credit department has utilized the classical credit scoring card model to establish the credit rating system of the bank.At the beginning of 2019,Bank A launched a product of personal credit loan.Due to the limitations of the traditional score card method,which relies on the subjective experience of credit experts,the efficiency and accuracy of credit rating system are largely affected.Machine learning methods have been widely applied in practice and achieved excellent results based on their superior learning and rapid self-adjusting abilities.In this thesis,we apply several machine learning methods to improve this practical problem faced by the credit department of Bank A.In this thesis,the author selected three classical and widely-used machine learning methods in the field of personal credit rating: Decision Tree Model,Logistic Regression Model and Support Vector Machine(SVM)Model,and make a comprehensive numerical experiment to compare with the traditional credit scoring card model.The results prove that the three machine learning methods are superior to the traditional credit scoring card method both in terms of efficiency and accuracy of classification.This indicates that the machine learning methods have a great potential to be applied in the actual work of personal credit rating in Bank A.Moreover,we put forward proposals for practical application combining with the characteristics selection of Bank A's business of personal credit loan.In this thesis,the numerical data set contains more than 20,000 samples of the personal credit loan of Bank A.First,a qualitative and quantitative analysis is used to preprocess the samples.Especially,information gain ratio is used to select the characteristics,which improves the efficiency and accuracy of the machine learning models.Then the samples are shown by some appropriate charts to make the statistical results more intuitive and persuasive.Finally,the samples are tested by each model,and some measure criteria are selected to evaluate the results.The contributions of this thesis can be summarized as follows.First,it provides a good guidance for Bank A to develop a unique but refined way to establish the personal credit rating system.Second,compared with the traditional credit scoring card system,machine learning system avoids the influence of subjective judgment of credit experts and builds a standardized operation system.Third,this new system effectively enhances the risk ability of the bank while greatly reduces its operational cost.It's helpful for the sustainable development of Bank A.