“Application of Graph Neural Networks in Financial Fraud Detection”
“Application of Graph Neural Networks in Financial Fraud Detection”
AliReza Solaimani1 Ehsan Soleimani Dehkordi2
1) Student of Computer Engineering, National University of Skills, Shahrekord, Chaharmahal and Bakhtiari, Iran”
2) PhD student of Computer Engineering, Azad University, Meybod Branch, Iran
IOI Code
:
AAHD-ABEFEA
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
Abstract
This paper presents a comprehensive review of the growing application of Graph Neural Networks (GNNs) in the critical domain of financial fraud detection. Given the graph-structured nature of transactional data and the complex relationships between entities (such as users, accounts, and transactions), GNNs have emerged as a powerful tool for identifying hidden fraud patterns. The primary objective of this research is to provide an overview of the latest advancements, key GNN models, and existing challenges in this field. The research methodology is based on a systematic review of the existing scientific literature and an analysis of articles using GNNs in various types of financial fraud, including banking transactions, credit card fraud, and online trading. Preliminary results indicate that GNN models, especially those capable of understanding the heterogeneous features of graphs, have demonstrated superior performance compared to traditional methods in detecting suspicious cases. However, challenges such as scalability, model interpretability, and the need for high-quality labeled data still require further attention and research. This review provides a framework for researchers and practitioners in the financial security domain to take steps towards developing innovative solutions by gaining a deeper understanding of the capabilities and limitations of GNNs.
Keywords : graph neural networks financial fraud detection data mining machine learning heterogeneous graphs