Development of Novel Inflation-Aware AI Framework for Credit Risk Assessment under Economic Stress

Authors:
Carla Popolizio, Rejwan Bin Sulaiman, Maruf Farhan, Naser Abosata

Addresses:
Department of Computer Science and Technology, Northumbria University, Newcastle upon Tyne, London, United Kingdom. Department of Science and Engineering, School of Computing, University of Derby, Derby, England, United Kingdom.

Abstract:

Traditional credit scoring models use historical financial data to estimate borrower risk, but they rarely consider how borrower affordability might change under inflation or economic instability. This limits lenders, as borrower default risk may rise even when underlying credit quality remains the same. To address this gap, this study created and tested an Inflation-Aware AI Credit Risk Decision Support Framework that uses supervised machine learning and inflation stress testing to help lenders make lending decisions during economic uncertainty. Using a Design Science Research methodology, researchers built and evaluated three supervised machine learning algorithms (Logistic Regression, Random Forest, and Gradient Boosting) on the Give Me Some Credit dataset, which contains 150,000 borrower records. Model performance was measured using accuracy, precision, recall, F1-score, and ROC-AUC, with stratified 5-fold cross-validation and McNemar’s test used to validate performance differences. Gradient Boosting performed best and was later chosen for deployment in an interactive Streamlit application. The approach predicts borrower default probability, categorises debtors, and assesses how default risk changes under mild, moderate, and severe inflation scenarios by simulating income losses and increases in debt ratio and credit usage. To support decision-making, the application includes a simple dashboard that displays default probability, loan recommendations, borrower-level default risk, and inflation-scenario comparisons. The results show that applying inflation stress testing to credit risk assessments adds more value beyond traditional methods by identifying borrowers whose risk profile worsens under uncertain scenarios. The suggested approach therefore provides a useful tool that shows how AI and economic scenario analysis can further enhance evaluations in real financial contexts.

Keywords: Credit Risk Assessment; Decision Support Framework; Machine Learning; Random Forest; Borrower Default Risk; Gradient Boosting; AI Framework.

Received on: 23/06/2025, Revised on: 28/09/2025, Accepted on: 01/12/2025, Published on: 05/06/2026

DOI: 10.69888/FTSTPL.2026.000813

FMDB Transactions on Sustainable Technoprise Letters, 2026 Vol. 4 No. 2, Pages: 104-126

  • Views : 63
  • Downloads : 8
Download PDF