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

2026-02-27

AI Summary

This study addresses the performance limitations of graph neural networks (GNNs) in anti-money laundering (AML) tasks, which often stem from suboptimal training strategies—particularly when applied to highly imbalanced blockchain transaction data. The authors systematically evaluate the impact of various weight initialization and normalization schemes on GCN, GAT, and GraphSAGE using the Elliptic Bitcoin dataset, revealing for the first time a pronounced architecture-dependent sensitivity to training configurations. Through reproducible ablation studies incorporating temporal data splits, controlled random seeds, and comprehensive hyperparameter tuning, they find that GraphSAGE achieves optimal performance with Xavier initialization alone, GAT requires a combination of Xavier initialization and GraphNorm, while GCN remains largely insensitive to these choices. These findings offer clear, architecture-specific guidance for deploying GNNs in AML applications.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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.

Problem

• Graph Neural Networks
• Anomaly Detection
• Anti-Money Laundering
• Initialisation
• Normalisation

Innovation

• Graph Neural Networks
• Normalization
• Initialization
• Anti-Money Laundering
• Ablation Study