A Hybrid Method for Online Loan Default Prediction Using Machine Learning Techniques

Predicting loan credit risk is a fundamental issue in banking activities. Since this process requires the analysis of a large volume of data with various variables, traditional approaches are time-consuming and inefficient. As a result, in recent years, there has been a growing interest in using machine learning techniques, as well as feature selection techniques, which have become essential tools in this field. However, due to the imbalanced nature of data in this domain, machine learning techniques alone often do not perform well. In recent years, researchers in this field have proposed solutions for data balancing, feature selection, and machine learning models to address these challenges. Since most studies in the literature have addressed the reduction of features and data balancing separately, this research presents a combined approach using machine learning models alongside balancing methods to increase prediction accuracy, as well as feature selection algorithms to reduce the credit scoring time in online lending. This study was conducted on a dataset consisting of 8289 records from one of Iran's banks. The results of this study indicate that the Random Forest and XGBoost models, in general, performed the best compared to other models.