Research On Credit Evaluation Of Bank Personal Credit Loan Based On SMOTE-Logistic Regression Algorithm

With social development,the consumption pattern of people has transformed progressively,and the volume of individual credit loan matter of major banks has grown annually.How to stem the credit risk of individual credit loan has become one of the key points of banking practitioners by degrees.As an important index to measure the customer's personal credit,credit score is the intuitive embodiment of the customer's personal integrity,and it is also the approval criteria of bank lending and the momentous proof to reduce the lending risk.Therefore,it is quite significant for banks to abate decision-making hazards,to enhance comprehensive profits and to bring about the sound expansion of individual credit loan operation.The significant points of the research rest with the subsequent several aspects.This paper preprocesses and analyzes personal credit information.In this article,it takes the customer's credit information of the item named “Give Me Some Gredit”as research data.Using average value to fill in the disappearing data,and utilizing the practical significance or the box diagram to resolve the abnormal data.Histogram,kernel density estimation and thermodynamic diagram are used to visualize the correlation between feature variables,and collinear feature variables with large correlation coefficient are eliminated.This paper uses the Logistic regression model improved by SMOTE algorithm to improve the accuracy of personal credit assessment.In the light of the shortcoming of Logistic regression algorithm,this article proposes improvements.Through the reference of SMOTE algorithm,in order to determine the improvement based on SMOTE algorithm.After balancing and adjusting the category structure of sample data,it can provide data support for the subsequent construction SMOTE-Logistic regression model.By selecting the appropriate evaluation index,it is capable of providing the reference for evaluating the comprehensive effect of the model.Through the comparison with the single model such as the KNN? the random forest and the combination model such as SMOTE-KNN ? SMOTE-random forest,it is demonstrated that the SMOTE algorithm can ameliorate the correctness of the Logistic regression pattern.This article structures the credit readiness scorecard model of personal credit loan,and applies the system to the estimation and the factual data.In this thesis,the characteristic variables of the samples are screened by the method of equal depth segmentation.This paper calculates and analyzes the WOE value and IV value under each variable grouping,so as to hold back the fundamental characteristic variables.By introducing and transforming the SMOTE-Logistic regression algorithm,the intercept term value and the regression parameter value of each characteristic variable are obtained,and the individual basic credit score and the group score of each characteristic variable are obtained.That is,the standard credit score card model is constructed.In this paper,the corresponding credit score is obtained by inputting the randomly selected customer credit information into the personal credit score card model.By dividing the credit score,we can get four credit grades.Based on different credit rating,this paper analyzes from two aspects: customer individual and bank.This research can give basis for the business approval,and offer new research ideas for bank personal credit loan credit evaluation by giving targeted and reasonable suggestions.