Forecasting nonperforming loans using machine learning

Mohammad Abdullah, Mohammad Ashraful Ferdous Chowdhury, Ajim Uddin, Syed Moudud-Ul-Huq

Research output: Contribution to journalArticlepeer-review

4 Scopus citations

Abstract

Nonperforming loans play a critical role in financial institutions' overall performance and can be controlled by forecasting the probable nonperforming loans. This paper employs a series of machine learning techniques to forecast bank nonperforming loans on emerging countries' financial institutions. Using quarterly cross-sectional data of 322 banks from 15 emerging countries, this study finds that advanced machine learning-based models outperform simple linear techniques in forecasting bank nonperforming loans. Among all 14 linear and nonlinear models, the random forest model outperforms other models. It achieves a 76.10% accuracy in forecasting nonperforming loans. The result is robust in different performance metrics. The variable importance analysis reveals that bank diversification is the most critical determinant for future nonperforming loans of a bank. Additionally, this study revealed that macroeconomic factors are less prominent in predicting nonperforming loans compared with bank-specific factors.

Original languageEnglish (US)
Pages (from-to)1664-1689
Number of pages26
JournalJournal of Forecasting
Volume42
Issue number7
DOIs
StatePublished - Nov 2023

All Science Journal Classification (ASJC) codes

  • Modeling and Simulation
  • Economics and Econometrics
  • Computer Science Applications
  • Strategy and Management
  • Statistics, Probability and Uncertainty
  • Management Science and Operations Research

Keywords

  • bagged CART
  • banking
  • forecasting
  • machine learning
  • nonperforming loans (NPLs)

Fingerprint

Dive into the research topics of 'Forecasting nonperforming loans using machine learning'. Together they form a unique fingerprint.

Cite this