Normalisation and Initialisation Strategies for Graph Neural Networks in Blockchain Anomaly Detection

TL;DR

This paper systematically studies how initialisation and normalisation strategies affect the performance of different GNN architectures in blockchain anti-money laundering tasks, providing practical guidance for real-world applications.

Contribution

It offers the first comprehensive analysis of architecture-specific initialisation and normalisation effects on GNNs for AML detection, with practical recommendations.

Findings

GraphSAGE performs best with Xavier initialisation.

GAT benefits from GraphNorm and Xavier initialisation.

GCN shows limited sensitivity to these strategies.

Abstract

Graph neural networks (GNNs) offer a principled approach to financial fraud detection by jointly learning from node features and transaction graph topology. However, their effectiveness on real-world anti-money laundering (AML) benchmarks depends critically on training practices such as specifically weight initialisation and normalisation that remain underexplored. We present a systematic ablation of initialisation and normalisation strategies across three GNN architectures (GCN, GAT, and GraphSAGE) on the Elliptic Bitcoin dataset. Our experiments reveal that initialisation and normalisation are architecture-dependent: GraphSAGE achieves the strongest performance with Xavier initialisation alone, GAT benefits most from combining GraphNorm with Xavier initialisation, while GCN shows limited sensitivity to these modifications. These findings offer practical, architecture-specific guidance for deploying GNNs in AML pipelines for datasets with severe class imbalance. We release a reproducible experimental framework with temporal data splits, seeded runs, and full ablation results.