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 show 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 80x lower computational 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.Editorial revision. The prose was converted to an impersonal, formal register (single-author paper). No data, numbers, claims, tables, or references were changed from the previous version.
Karl J. Mollan Neyra (Thu,) studied this question.