Comparison of different individual credit risk assessment models

Abstract

Personal credit risk is increasing in the background of continuous expansion of bank credit business. Previous researchers use many algorithms of machine learning to assess personal credit risk, while these models differ in real scenarios and accuracy. Based on this situation, this research first analyzes the influencing factors of individual credit risk through searching data, and then shows the relationship between them in figures. Meanwhile, this research compares three machine learning models in the predicting the accuracy. These models are Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Logistic Regression (LR). Learning from the previous studies on individual credit risk assessment model, this research compares the final average value of Area Under Curve (AUC), which is obtained by calculating AUC one and AUC two. The results show that XGBoost has better performance than the other models for a high AUC value. This research provides an idea for banks to select and individual credit risk models.