Title: Comparative analysis of machine learning models for predicting online loan defaults in Nigeria

Authors: Samuel Gbenga Faluyi; Peter Adebayo Idowu; Gbenga Oyewole Ogunsanwo; Olumuyiwa Bamidele Alaba

Addresses: Department of Computer Science, Tai-Solarin University of Education, Ijagun, Ijebu-Ode, Nigeria ' Department of Computer Science and Engineering, Obafemi Awolowo University, Ile-Ife, Nigeria ' Department of Computer Science, Tai-Solarin University of Education, Ijagun, Ijebu-Ode, Nigeria ' Department of Computer Science, Tai-Solarin University of Education, Ijagun, Ijebu-Ode, Nigeria

Abstract: The purpose of this study is to carry out a comparative analysis of machine learning models developed for online loan defaulters' prediction. This paper adopted four machine learning approaches, which include random forest, XGBoost, CatBoost and AdaBoost algorithms for predictions of online loan defaulters. The dataset used consists of loan transactions carried out within 2017 in Nigeria's online loan company (super lender), obtained from the Zindi dataset repository. The data comprises 4,346 instances and 21 attributes. The data was cleaned and pre-processed before the models were fitted using Python 3.8 Jupyter Notebook environment for simulation and were validated using standard performance metrics, which include the accuracy measure, F1-score, precision, recall, and AUC-ROC score. Based on the results of the study it was identified that CatBoost outperformed other algorithms based on comparison of the performance. Thus, the paper concluded that CatBoost and AdaBoost can be incorporated into online loan systems to determine borrowers' creditworthiness.

Keywords: online loan default; machine learning; ML; performance metric; creditworthiness; Nigeria.

DOI: 10.1504/IJAISC.2024.145626

International Journal of Artificial Intelligence and Soft Computing, 2024 Vol.8 No.3, pp.238 - 256

Received: 10 Jul 2023
Accepted: 14 May 2024
Published online: 09 Apr 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article