Personal Credit Evaluation Model Based On SVM And XGBoost

With the development of economy and society,credit has become an important part of daily life,so we need scientific methods for credit evaluation.In the current era of big data,data has the characteristics of diversified structure and huge amount of data.It is difficult for traditional subjective credit evaluation methods to respond to rapid market changes.Therefore,various financial institutions need to use statistical methods to build model and realize credit scoring to reduce credit risk.This thesis proposes an improved weighted least squares support vector machine(WLS-SVM)based on the traditional support vector machine(SVM)and compares it with the integrated algorithm XGBoost.Weighted least squares support vector machine proposes to use weighted penalty factors to control the importance of different categories,and convert the inequality constraints of traditional support vector machines into equality constraints.We choose a real personal customer loan default data set to verify the effectiveness and feasibility of the algorithm.Process visual analysis,missing value processing(mean imputation),standardization processing(Z-score standardization),data imbalance processing(SMOTE)and feature selection(t test)for the data set,and finally establish a model and use the SHAP method to carry out features importance explanation.The results show that the weighted least squares support vector machine is better than the traditional support vector machine in classification accuracy,precision,recall and AUC value.The recall rate is one to two percentage points higher than XGBoost,and the AUC value is slightly lower than XGBoost.In general,in terms of credit risk assessment,the weighted least squares support vector machine and XGBoost used in this paper have a good classification effect,which has guiding significance for the credit risk control of financial institutions.