Construction of a financial default risk prediction model based on the LightGBM algorithm

Abstract

Abstract The construction of a financial risk prediction model has become the need of the hour due to long-term and short-term violations in the financial market. To reduce the default risk of peer-to-peer (P2P) companies and promote the healthy and sustainable development of the P2P industry, this article uses a model based on the LightGBM (Light Gradient Boosting Machine) algorithm to analyze a large number of sample data from Renrendai, which is a representative platform of the P2P industry. This article explores the base LightGBM model along with the integration of linear blending to build an optimal default risk identification model. The proposed approach is applicable for a large number of multi-dimensional data samples. The results show that the prediction accuracy rate of the LightGBM algorithm model on the test set reaches 80.25%, which can accurately identify more than 80% of users, and the model has the best prediction performance in terms of different performance evaluation indicators. The integration of LightGBM and the linear blending approach yield a precision value of 91.36%, a recall of 75.90%, and an accuracy of 84.36%. The established LightGBM algorithm can efficiently identify the default of the loan business on the P2P platform compared to the traditional machine learning models, such as logistic regression and support vector machine. For a large number of multi-dimensional data samples, the LightGBM algorithm can effectively judge the default risk of users on P2P platforms.