Â鶹APP

Explainable AI-Driven Decision Support for Robust Fraud Detection and Risk Mitigation

• Ruiqing Xu, Stuart Ph.D. student

Abstract:

This research addresses financial fraud detection in e-commerce by integrating high-performance models (XGBoost, SVM, and Logistic Regression) with advanced XAI techniques (SHAP, LIME, Anchors, LORE, and an ensemble XAI). It focuses on minimizing financial risk—both from false positives that inconvenience legitimate customers and from false negatives that result in direct losses, regulatory penalties, and reputational damage. Incremental learning ensures adaptation to shifting fraud patterns, reflecting the dynamic nature of financial crime. Rigorous validation, including perturbation analysis, sliding-window cross-validation, and adversarial training, strengthens robustness against noise, bias, and concept drift. By leveraging interpretable insights, this system supports critical risk-management objectives, enhances compliance, and promotes transparency for auditors and stakeholders. Ultimately, this scalable, explainable AI solution mitigates financial risk while preserving trust in real-time e-commerce transactions.

All Illinois Tech faculty, students, and staff are invited to attend.

The Friday Research Presentations series showcases ongoing academic research projects conducted by Stuart School of Business faculty and students, as well as guest presentations by Illinois Tech colleagues, business professionals, and faculty from other leading business schools.