Research And Implementation Of Personal Loan Credit Evaluation Model

The credit service of commercial banks is the most important source of income for commercial banks.Banks lend to customers with poor repayment ability or low credit.This will make it difficult for banks to recover the money they lent,which will cause losses to banks.Through credit evaluation of personal loans,commercial banks can predict whether users will default on repayments and control lending to reduce the losses caused by customer default.Personal loan credit evaluation is essentially a classification problem,distinguishing between customers with good credit and customers with poor credit.Scholars have been looking for personal credit evaluation models with good accuracy and stability.Scholars have used statistical methods(including linear regression,discriminant analysis,logical regression,decision tree,etc.)non-statistical methods,expert systems,neural networks,genetic algorithms,and combination methods.With the development of commercial banking business,the amount of bank credit data is increasing.In order to make personal credit assessment more accurate and efficient,this paper proposes a credit evaluation model for personal loans based on Light GBM.First of all,the data are analyzed and processed,and K-Nearest Neighbor algorithm is used to fill the missing value when the data is missing.Then,the customers of credit evaluation are grouped by K-means clustering algorithm.Group customers according to their monthly income and debt ratio.Customers are divided into four groups:low-income and low-debt groups,low-income and high-debt groups,high-income and low-debt groups,and high-income and high-debt groups.Finally,Light GBM is used to train customer group data to obtain a personal loan credit evaluation model.This model can more accurately predict whether users will default or not.In view of the small number of customer default samples,this paper uses clustering algorithm to group customers.Through model empirical analysis,using grouping strategy is better than non-grouping strategy.Grouping strategy model can more accurately predict whether customers default or not.Light GBM algorithm has the characteristics of faster training efficiency,support for parallel learning,and can handle large-scale data.Through the empirical analysis of credit evaluation model and comparison of XGBoost-based evaluation model,it is concluded that Light GBM-based evaluation model is more accurate and efficient in predicting whether users default or not.In order to make it easier to conduct personal loan credit assessments,using the evaluation model based on Light GBM,a visual personal credit evaluation system is designed and implemented.The system uses a front-end separation framework,the front end uses Vue,and the back end uses the Spring Boot framework.The system interface is simple and elegant,and the operation is simple.The system can automatically call the Light GBM evaluation model for personal credit evaluation,which is quick and convenient.The system also performs statistical analysis on customer credit distribution,generates charts,and the data can be visually displayed to users.The system is capable of managing customer data and automating personal credit assessments to make personal credit assessments more convenient.