Dynamic Evaluation Model of the Default Risk of Online Loan Borrowers Based on K-means and SVM

Abstract

Objective: With the development of Internet finance, the scale of peer-to-peer (P2P) online lending platforms have rapidly expanded. Additionally, the phenomenon of losses among online lending platforms and the problem of default by borrowers have emerged, greatly restricting the healthy development of online lending platforms. Therefore, it is necessary to dynamically set the credit rating of the borrowers according to the performance of the borrowers, and establish a default risk evaluation model for the borrowers of the online lending platform to promote the healthy development of the online lending platform. Method: This paper uses web crawler technology to obtain borrower information as the sample data, selects 17 core variables as explanatory variables, and utilizes project status as the target variable. First, according to the performance of the borrower, we use K-means to cluster, obtain a dynamic credit rating by calculation and reset the rating to obtain new borrower information. Second, we determine the optimal parameters of the support vector machine algorithm through cross-validation and establish the best evaluation model for online loan borrowers' default risk. Finally, we conduct experimental verification. Results: The classification accuracy of the proposed algorithm is better than that of decision trees and random forest, and the classification effect is the strongest. Conclusion: The experimental results show that the model has good stability and generalizability, and the research results provide dynamic decision support for early warnings of online lending platform risk and risk prevention and control; they can thus help promote the healthy development of online lending platforms.