Graph Learning for Trade Based Money Laundering Using Graph Neural Networks to Uncover Suspicious Trade Routes and Payment Patterns Indicative of Money Laundering

Abstract

Trade-based money laundering (TBML) poses a significant challenge to global financial systems, exploiting the complexity of international trade to disguise illicit funds through manipulated transactions. Traditional detection methods, which analyze transactions in isolation, fail to capture the networked nature of TBML, resulting in high false-positive rates and inefficiencies. This study addresses this critical gap by leveraging graph neural networks (GNNs) to model trade networks as interconnected graphs, where nodes represent entities (e.g., exporters, importers) and edges capture transactional relationships. The primary objective was to develop a GNN-based framework capable of identifying suspicious trade routes and payment patterns by analyzing structural and transactional anomalies. Using a dataset of 1.2 million trade records (2015–2022) from UN Comtrade, WCO, and FATF typologies, we constructed multi-relational graphs and applied Graph Convolutional Networks (GCNs) and Graph Attention Networks (GATs) for anomaly detection. Key findings revealed statistically significant discriminators of TBML, including transaction value (t = -48.26, p < 0.001), currency fluctuations (t = -28.71, p < 0.001), and irregular payment terms (χ² = 15.51, p = 0.001). Logistic regression further confirmed transaction value (coef. = 0.000081, p < 0.001) as a robust predictor. The GNN framework achieved superior detection accuracy compared to traditional methods (AUC-ROC = 0.92), reducing false positives by 30%. These results demonstrate that graph learning enhances TBML detection by uncovering latent network structures, offering financial institutions and regulators a scalable, interpretable tool. This research bridges the gap between theoretical graph-based analytics and practical anti-money laundering (AML) applications, providing a foundation for future AI-driven compliance solutions