XAI-driven Approaches to Detect Financial Frauds and Mitigate Risk

Constructively, in modern finance, the two most important elements are fraud detection and risk management, to be able to keep public faith and be able to abide by the regulators. This chapter is in recognition of advanced AI methods – machine learning, deep learning, and hybrid modeling, revealing anomalous transactions in volume financial data. In addressing imbalanced data, dynamic transaction patterns, and privacy, we work on the 2023 Credit Card Fraud Detection Dataset, containing more than 550,000 anonymized records from European cardholders. We intermingle supervised and unsupervised algorithms XGBoost, Isolation Forest, DBSCAN, K-means, and Logistic Regression to detect both established and emerging fraud. While Logistic Regression achieves around 96.36% performance, XGBoost provides up to 99.88% accuracy, illustrating the trade-offs between performance level and transparency.