PREF-Gate: Provenance-Constrained Relational Evidence Fusion with Validation-Gated Selection for Graph Fraud Detection

2026-07-13

AI Summary

This work addresses the vulnerability of neighborhood evidence to label contamination from validation or test sets in graph-based fraud detection. To mitigate this issue, the authors propose an auditable decision framework that coordinates an unlabeled-context expert with an evidence expert relying solely on training labels, coupled with a limited-validation gating mechanism that fixes expert selection and decision thresholds prior to testing. Innovatively grounded in a label-provenance constraint principle, the framework enforces explicit contracts on label origins, employs a constrained selection strategy, enables failure attribution, and supports audit-budget evaluation, thereby ensuring adaptive yet verifiable decision conditions. Integrating techniques such as one-hop mean aggregation, feature residuals, and degree descriptors, the method achieves average AUPRC scores of 0.9085, 0.8104, and 0.8913 on Amazon, YelpChi, and TFinance, respectively, demonstrating that label-derived evidence is effective only when supported by validation.

Abstract

Relational fraud detection can exploit both label-free graph context and label-derived neighborhood evidence, but these two information sources obey different validity conditions. In particular, neighborhood risk becomes invalid when a queried node's own label, or any validation or test label, enters its construction. We formulate this issue as provenance-constrained relational evidence use and present PREF-Gate, an auditable decision framework with two fixed experts and a finite validation gate. The context expert uses attributes, one-hop means, feature residuals, and degree descriptors without labels. The evidence expert adds self-excluded, training-label-only neighborhood risk and empirical-Bayes summaries that expose support, uncertainty, availability, and shrinkage. Before test inference, the gate selects either expert or one of three pre-specified probability mixtures and fixes the decision threshold. On Amazon, YelpChi, and TFinance, using five identical stratified splits and 14 same-protocol methods, PREF-Gate obtains mean AUPRC values of 0.9085, 0.8104, and 0.8913. It selects the label-free expert on all Amazon and YelpChi splits and an evidence mixture on all TFinance splits. Thus, the main result is conditional rather than universal: label-derived relational evidence is useful only where held-out validation supports it. The framework couples competitive ranking performance with an explicit label-provenance contract, finite selection policy, failure accounting, and review-budget evaluation, providing an auditable knowledge-based decision pipeline for graph fraud detection.

Problem

• graph fraud detection
• relational evidence
• label provenance
• evidence validity
• neighborhood risk

Innovation

• provenance-constrained
• validation-gated selection
• relational evidence fusion
• graph fraud detection
• auditable decision framework