Enhancing Credit Card Fraud Detection with Explainable AI and Model Comparison
Anushka Srivastava, Soumya Singh, Akshay Kumar, T. Ragupathi · 2025
With the rapid growth of digital financial transactions, detecting fraudulent activities has become a major challenge for financial institutions. Traditional machine learning models often struggle with imbalanced datasets and the complexity of fraud patterns. In this paper, we suggest a fraud detection model using XG Boost that effectively captures the transactional relationships between users, merchants, and transactions. Additionally, we incorporate Explainable AI (XAI) techniques using Shapley Additive Explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) to improve transparency, making fraud predictions more interpretable. Our results demonstrate that the XG Boost-based approach outperforms traditional classifiers such as Random Forest, SVM, and Logistic Regression in identifying fraudulent transactions. The integration of SHAP and LIME-based feature attribution further enhances model credibility, ensuring regulatory compliance and fostering trust in financial applications.