Economics Research

Predicting Loan Default Behavior Using Machine Learning Models: A Case Study of Resalat Bank

Articles in Press

Document Type : Research Paper

Author

Fakhreddin Fakhrhosseini

Department of Accounting, TONIAU, Islamic Azad University, Tonekabon, Iran

https://doi.org/10.22054/joer.2026.93680.1347
Abstract
In recent years, the surge in credit risk among customers and the escalation of economic crises have contributed to rising loan default rates in the banking sector. In this context, identifying high-risk customers has become a top priority for banks in credit risk management. This research employs real customer data from Resalat Bank, spanning from May 22, 2020, to March 18, 2025, to predict loan default behavior using machine learning models. The dataset was processed based on demographic and credit history attributes, including age, gender, loan amount, number of installments, and payment delays. To evaluate performance, Logistic Regression models with L1 and L2 regularization, Neural Networks, Random Forest, and XGBoost were employed. Furthermore, the Synthetic Minority Over-sampling Technique (SMOTE) was utilized to address data class imbalance. The results indicate that decision tree-based models, particularly XGBoost and Random Forest, outperform Logistic Regression and Neural Networks in terms of Recall, Precision, and F1-score metrics. These findings suggest that combining relevant data with robust algorithms designed for imbalanced datasets can significantly improve the accuracy of default risk prediction.

Keywords

Credit Loans Risk
Financial Default
Machine Learning
Subjects
Economics Research

Articles in Press, Accepted Manuscript
Available Online from 26 September 2026

  • Receive Date 20 June 2026
  • Revise Date 17 August 2026
  • Accept Date 26 September 2026

Home

Editorial Board

Guide for Authors

Submit Manuscript

Contact Us