What question did this study set out to answer?

GRAFID aims to help financial institutions assess the value of graph-based fraud detection methods versus their computational costs.

April 12, 2026Open Access

GRAFID: A Cost-Aware Decision Framework for Graph-Enhanced Fraud Detection in Payment Systems

Key Points

GRAFID aims to help financial institutions assess the value of graph-based fraud detection methods versus their computational costs.
Developed a decision framework for evaluating graph-enhanced fraud detection.
Conducted experiments using two public datasets: IEEE-CIS and European Credit Card.
Compared performance of XGBoost and GraphSAGE across various model configurations.
On the IEEE-CIS dataset, XGBoost achieved an AUPRC of 0.508, outperforming GraphSAGE's 0.411.
XGBoost's performance was achieved at approximately one-sixtieth the cost of GraphSAGE.
On the feature-sparse Credit Card dataset, both methods performed identically.

Abstract

We introduce GRAFID, a decision framework that helps financial institutions determine whether graph-based methods justify their computational overhead for transaction fraud detection. Through twenty-plus model configurations across two public datasets (IEEE-CIS and European Credit Card), we demonstrate that the value of graph structure depends primarily on feature richness. On the feature-rich IEEE-CIS dataset, XGBoost achieves AUPRC of 0.508, outperforming GraphSAGE (0.411) at roughly one-sixtieth the cost. On the feature-sparse Credit Card dataset, the two approaches perform identically. GRAFID provides three actionable indicators: Feature Richness Index, Graph Signal Gain, and Cost-Effectiveness Ratio, enabling practitioners to make informed investment decisions before committing to graph infrastructure.

GRAFID: A Cost-Aware Decision Framework for Graph-Enhanced Fraud Detection in Payment Systems

Key Points

Abstract

Cite This Study