Detecting Fraudulent Transactions for Different Patterns in Financial Networks Using Layer Weigthed GCN

Detecting Fraudulent Transactions for Different Patterns in Financial Networks Using Layer Weigthed GCN

Abstract As digital financial transactions and fraudulent activities grow in complexity, ensuring financial system security becomes increasingly challenging. This paper introduces LayerWeighted-GCN (LWG), a novel Graph Neural Network (GNN) model for fraud detection, along with a synthetic dataset na...

Full description

Saved in:
Bibliographic Details
Main Authors: Shaziya Islam, Gagan Raj Gupta, Apu Chakraborty, Santosh Singh, Anisha Soni, Chhavi Patle
Format: Article
Language:English
Published: Springer Nature 2025-04-01
Series:Human-Centric Intelligent Systems
Subjects:
Graph Neural Network (GNN)
Graph Convolution Network (GCN)
GraphSage (GS)
Graph Attention Networks (GAT)
Layer Weighted GCN (LWG)
Online Access:https://doi.org/10.1007/s44230-025-00097-3
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract As digital financial transactions and fraudulent activities grow in complexity, ensuring financial system security becomes increasingly challenging. This paper introduces LayerWeighted-GCN (LWG), a novel Graph Neural Network (GNN) model for fraud detection, along with a synthetic dataset named SIFT (Synthetic Insights for Fraudulent Transactions). The proposed approach integrates adaptive layer weighting, enabling the model to dynamically adjust the contribution of different GNN layers to better capture intricate transaction patterns. Additionally, the flexible architecture of LayerWeighted-GCN enhances its ability to model complex financial relationships, improving fraud detection accuracy and robustness. To evaluate its effectiveness, we compare LayerWeighted-GCN against classical machine learning models and traditional GNNs on the SIFT dataset. Experimental results demonstrate that LWG achieves higher accuracy than existing GNN-based models, outperforming state-of-the-art methods in fraud detection. Moreover, LWG effectively reduces false positive rates, making it a more reliable and adaptable solution for identifying fraudulent transactions. These findings highlight the model’s efficiency and effectiveness, establishing it as a powerful tool for financial institutions to detect and prevent sophisticated fraudulent activities.
ISSN:2667-1336

Similar Items