Democratizing Fraud Detection: An Explainable XGBoost System with DeepSeek-Enhanced Analytics for Real-Time Credit Card Transaction Monitoring

Publication Details

Publication Date

10/08/2026

Volume / Issue

Vol 1, Issue 3 (2026)

Article No.

040

Journal

British Journal of Contemporary Research

DOI

https://doi.org/10.67693/BJCR-P6K6WXGM

Received

28 Jul 2026

Views

34

Downloads

16

Abstract

Credit card fraud losses were $33.45 billion in 2022 and are anticipated to exceed $43 billion by 2028 [1]. Machine learning algorithms are highly accurate at detecting fraud; their "black-box" nature hinders adoption in regulated financial institutions [3]. We present an explainable fraud detection system combining XGBoost ensemble learning with DeepSeek AI-powered explanations. On the Kaggle Credit Card Fraud Detection dataset (284,807 transactions, 0.17% fraud rate) [4], our system achieves 99.96% accuracy, 89.09% precision, 94.23% recall, and 0.9979 ROC-AUC using SMOTE-balanced training. Key contributions include SHAP/LIME integration for model interpretability and an interactive Streamlit web application providing real-time predictions with human-readable AI explanations [5, 7]. Our system demonstrates that high-performance fraud detection can coexist with transparency, addressing critical gaps in trust and regulatory compliance

License

This article is published under the Creative Commons Attribution 4.0 International License . Free to read, share, and adapt with attribution.