Predicting Bank Loan Risk Using Incremental Learning

Abstract

Predicting loan default is a critical issue in the banking sector, given the financial risks it entails, which impact banks' performance and stability. This paper proposes incremental learning-based models for classifying bank loans and estimating their probability of default, comparing these models with the performance of traditional one-shot models. A set of machine learning models, including logistic regression, decision tree, random forest, XGBoost, SVM, and multilayer neural network (MLN), were tested on data comprising financial and behavioral information on bank customers. The results showed that the XGBoost model achieved the best performance in terms of accuracy and stability in the incremental learning model, reaching 0.9340.