Research on Graph Neural Network Algorithms for Financial Anomaly Detection

• Published in: 2023 International Conference on Networks, Communications and Intelligent Computing (NCIC) pp. 18 - 23
• Main Authors: Lu, Xin, Duan, Chengge, Xie, Anqi, Bao, Menghu, Wang, Minze, Zhong, QiLong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 17.11.2023
• Subjects: Computational modeling; DGraph; Feature extraction; Finance; Financial anomalies; Graph neural networks; MLP; Prediction algorithms; Predictive models; Technological innovation; Training
• DOI: 10.1109/NCIC61838.2023.00009

In recent years, advanced digital technologies like big data, blockchain, and artificial intelligence have rapidly developed. While digital fintech has played an increasingly important role in society, benefiting development of the financial industries, it has also escalated anomaly risks. There is an urgent need to apply new digital technologies to traditional finance for anomaly prevention. Although numerous anomaly detection algorithms exist, financial anomaly detection in complex contexts remains theoretical. Given this, we compared three graph neural networks (GCN, GraphSAGE, GAT) against MLP model for anomaly prediction on the DGraph dataset. and the experiments prove that the graph neural network results all outperform the fully connected network.