Shapley Value-Guided Adaptive Ensemble Learning for Explainable Financial Fraud Detection with U.S. Regulatory Compliance Validation

2026-04-14 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, quick

Summary

A new study addresses the $32 billion annual cost of financial crime in U.S. institutions by developing AI tools for fraud detection that meet regulatory transparency requirements. The research evaluates explanation quality using faithfulness (sufficiency and comprehensiveness at k=5, 10, 15) and stability (Kendall's W across 30 bootstrap samples). XGBoost with TreeExplainer demonstrated near-perfect stability (W=0.9912), significantly outperforming LSTM with DeepExplainer (W=0.4962). The study introduces the SHAP-Guided Adaptive Ensemble (SGAE), which dynamically adjusts ensemble weights based on SHAP attribution agreement, achieving the highest AUC-ROC (0.8837 held-out; 0.9245 cross-validation). Additionally, a comprehensive evaluation of LSTM, Transformer, and GNN-GraphSAGE on the 590,540-transaction IEEE-CIS dataset showed GNN-GraphSAGE achieving AUC-ROC 0.9248 and F1=0.6013, with all results validated against OCC, SR 11-7, and BSA-AML compliance requirements.

Key takeaway

For AI Engineers developing financial fraud detection systems, this research demonstrates that explainable models can achieve high performance while satisfying U.S. regulatory compliance. You should prioritize models like XGBoost with TreeExplainer for their explanation stability and consider integrating the SHAP-Guided Adaptive Ensemble (SGAE) to enhance both accuracy and auditability, ensuring your solutions meet OCC Bulletin 2011-12 and Federal Reserve SR 11-7 requirements.

Key insights

Explainable AI models can achieve high fraud detection accuracy while meeting stringent U.S. financial regulations.

Principles

• Explanation stability is crucial for regulatory compliance.
• Dynamic ensemble weighting improves fraud detection.
• SHAP values enhance model transparency.

Method

The SHAP-Guided Adaptive Ensemble (SGAE) dynamically adjusts per-transaction ensemble weights based on SHAP attribution agreement to improve fraud detection and explainability.

In practice

• Use XGBoost with TreeExplainer for high stability.
• Implement SGAE for improved AUC-ROC.
• Evaluate models against OCC, SR 11-7, BSA-AML.

Topics

• Financial Fraud Detection
• Explainable AI
• Shapley Values
• Adaptive Ensemble Learning
• Regulatory Compliance

Best for: AI Engineer, Research Scientist, CTO, AI Scientist, Machine Learning Engineer, Legal Professional

Related on AIssential

• XAI-SOH-FL: Enhancing SOH-FL with Adaptive Aggregation and Explainable AI for Intrusion Detection in Heterogeneous IoT — XAI-SOH-FL improves federated intrusion detection in IoT via adaptive aggregation and SHAP-based explainability.
• Beyond Defensive Reporting: Machine Learning for Active Anti-Money Laundering Control in Insurance — Machine learning, particularly with fraud data, can effectively detect insurance money laundering pre-payout.
• Classifying by Proxy: Explainable and Reproducible Ensemble of Proxy Tasks for Child Sexual Abuse Imagery Classification — An ensemble of Proxy Tasks enhances CSAI classification with improved explainability, reproducibility, and security.
• Statistical Inference and Learning for Shapley Additive Explanations (SHAP) — New methods enable statistical inference and confidence intervals for global SHAP values, addressing a critical gap in explainable AI.
• Explainable AI in Production: A Neuro-Symbolic Model for Real-Time Fraud Detection — Integrating explanations directly into model architecture enables real-time, deterministic, and low-latency explainability.
• Explainable AI based ensemble model for the identification of Schizophrenia prodromal phase — An XAI-integrated ensemble ML model accurately predicts Schizophrenia prodromal symptoms for early clinical intervention.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.