Financial Fraud Detection using Quantum Graph Neural Networks

TL;DR

The paper confronts fraud detection in finance, where traditional methods struggle with evolving fraudulent patterns. It introduces Quantum Graph Neural Networks (QGNNs) that encode transactions as graphs, use angle-encoded quantum states, and apply Variational Quantum Circuit layers to classify fraud. On a real dataset, QGNNs achieve $AUC=0.85$ and $94.5\%$ accuracy, outperforming a classical GraphSAGE baseline ($AUC=0.77$; $92.3\%$ accuracy), demonstrating the potential advantages of quantum-graphic representations for imbalanced data. The work highlights the role of topological data analysis and quantum processing in capturing complex relational patterns, suggesting practical gains in fraud detection and motivating further QC-enabled graph learning research.

Abstract

Financial fraud detection is essential for preventing significant financial losses and maintaining the reputation of financial institutions. However, conventional methods of detecting financial fraud have limited effectiveness, necessitating the need for new approaches to improve detection rates. In this paper, we propose a novel approach for detecting financial fraud using Quantum Graph Neural Networks (QGNNs). QGNNs are a type of neural network that can process graph-structured data and leverage the power of Quantum Computing (QC) to perform computations more efficiently than classical neural networks. Our approach uses Variational Quantum Circuits (VQC) to enhance the performance of the QGNN. In order to evaluate the efficiency of our proposed method, we compared the performance of QGNNs to Classical Graph Neural Networks using a real-world financial fraud detection dataset. The results of our experiments showed that QGNNs achieved an AUC of $0.85$, which outperformed classical GNNs. Our research highlights the potential of QGNNs and suggests that QGNNs are a promising new approach for improving financial fraud detection.

Figures (12)

• Figure 3: A directed graph. The arrows indicate the direction of flow. The graph consists of nodes labelled $A$, $B$, $C$, $D$, and $E$.
• Figure 7: Basic structure of Feed-Forward GNNs. The network consists of an input layer with $3$ nodes, a hidden layer with $4$ nodes, and an output layer.
• Figure 8: Architecture of Graph Convolutional Neural Network. The network includes a convolution layer followed by a ReLU activation function, a subsequent convolution layer with another ReLU activation function, an intermediate dense layer, and finally, a sigmoid activation function before the output layer.
• Figure 9: Structure of a Graph Recurrent Neural Network. The GRNN consists of an input layer with $4$ nodes, a hidden layer with $5$ nodes, and an additional layer with $3$ nodes. The architecture is completed with an output layer.
• Figure 10: Architecture of the GraphSage. It comprises successive GraphSAGE layers, each followed by a ReLU activation function. An intermediate dense layer bridges the structure, culminating in a sigmoid activation function in the output layer.