Explainable AI for credit card fraud detection: Bridging the gap between accuracy and interpretability

Ojo, Innocent Paul and Tomy, Ashna (2025) Explainable AI for credit card fraud detection: Bridging the gap between accuracy and interpretability. World Journal of Advanced Research and Reviews, 25 (2). pp. 1246-1256. ISSN 2581-9615

Credit card fraud poses a persistent threat to the financial sector, demanding robust and transparent detection systems. This study aims to address the balance between accuracy and interpretability in fraud detection models by applying Explainable AI (XAI) techniques. Using a publicly available dataset from Kaggle, we explored multiple machine learning models, including Random Forest and Gradient Boosting, to classify fraudulent transactions. Given the imbalanced nature of the dataset, SMOTE was used for oversampling to ensure model fairness. The XAI techniques SHAP and LIME were employed to provide in-depth explanations of model predictions, enhancing transparency by highlighting key features influencing decisions. The results showed that both models achieved high detection performance, with Random Forest achieving a perfect accuracy score of 100%. Furthermore, XAI methods provided valuable insights into feature importance, fostering trust among stakeholders by improving model interpretability. These findings underscore the importance of integrating XAI into fraud detection systems to deliver reliable, transparent, and actionable insights for financial institutions. Future research should focus on scaling these models and expanding the use of XAI in real-time fraud detection frameworks.