FEGAT:一种特征增强的图注意力网络

FEGAT:A Feature-Enhanced Graph Attention Network

摘要

针对传统图神经网络很难处理大规模的复杂数据,对网络中节点和边的特征以及二者之间的联系难以完全利用的问题,提出一种特征增强的图注意力网络(FEGAT).该方法先将数据建模为节点与边的有向图,然后设计带有注意力机制的图神经网络模型,通过边特征加权与注意力系数动态聚合邻居节点信息,并采用多维边特征独立加权与并行拼接策略融合多维边特征,最后结合XGBoost算法完成节点分类.为验证模型的有效性,在反洗钱数据集AMLSim上进行实验,结果表明,该模型在不平衡数据场景下,精准率、召回率及F₁分数均得到显著提升.

Abstract

Aiming at the problems that traditional graph neural networks were difficult to handle large-scale complex data and it was difficult to fully utilize the features of nodes and edges in the network as well as the connections between them,we proposed a feature-enhanced graph attention network (FEGAT).The method first modelled the data as a directed graph of nodes and edges.Then,a graph neural network model with an attention mechanism was designed to dynamically aggregate neighbor node information through edge feature weighting and attention coefficients,and fused multi-dimensional edge features by adopting an independent weighting and parallel concatenation strategy for multi-dimensional edge features.Finally,it combined XGBoost algorithm to complete node classification.To verify the effectiveness of the model,experiments were carried out on the anti-money laundering AMLSim dataset.The results show that the model significantly improves precision,recall and F₁-score in imbalanced data scenarios.

关键词

图神经网络 / 注意力机制 / 图表示学习 / 洗钱检测

Key words

graph neural network / attention mechanism / graph representation learning / money laundering detection

参考文献

[1] Johnson J, Gupta A, Li F F. Image Generation from Scene Graphs[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ:IEEE, 2018:1219-1228.

[2] Li Y G, Yu R, Shahabi C, et al. Diffusion Convolutional Recurrent Neural Network:Data-Driven Traffic Forecasting[EB/OL].(2017-07-06)[2025-02-10]. https://arxiv.org/abs/1707.01926.

[3] Han J Q, Rong Y, Xu T Y, et al. Geometrically Equivariant Graph Neural Networks:A Survey[EB/OL].(2022-02-22)[2025-02-15]. https://arxiv.org/abs/2202.07230.

[4] Kipf T N, Welling M. Semi-supervised Classification with Graph Convolutional Networks[EB/OL].(2016-09-09)[2025-02-17]. https://arxiv.org/abs/1609.02907.

[5] Veličković P, Cucurull G, Casanova A, et al. Graph Attention Networks[EB/OL].(2017-10-30)[2025-01-23]. https://arxiv.org/abs/1710.10903.

[6] Chen J, Ma T F, Xiao C. Fast Learning with Graph Convolutional Networks via Importance Sampling[EB/OL].(2018-01-30)[2025-02-18]. https://arxiv.org/abs/1801.10247.

[7] Hamilton W L, Ying R, Leskovec J. Inductive Representation Learning on Large Graphs[J]. Advances in Neural Information Processing Systems, 2017, 30:1025-1035.

[8] Gong L Y, Cheng Q. Exploiting Edge Features for Graph Neural Networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway, NJ:IEEE, 2019:9211-9219.

[9] Wang Z M, Chen J, Chen H P. EGAT:Edge-Featured Graph Attention Network[C]//Artificial Neural Networks and Machine Learning.Berlin: Springer International Publishing, 2021:253-264.

[10] Jiang X D, Ji P S, Li S. CensNet:Convolution with Edge-Node Switching in Graph Neural Networks[C]// Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence.[S.l.]:IJCAI, 2019:2656-2662.

[11] Yang Y L, Li D S. NENN:Incorporate Node and Edge Features in Graph Neural Networks[C]// Asian Conference on Machine Learning.[S.l.]:PMLR, 2020:593-608.

[12] You J X, Du T Y, Leskovec J. ROLAND:Graph Learning Framework for Dynamic Graphs[C]// Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2022:2358-2366.

[13] Pareja A, Domeniconi G, Chen J, et al. Evolvegcn:Evolving Graph Convolutional Networks for Dynamic Graphs[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2020:5363-5370.

[14] Chen T Q, Guestrin C. XGBoost:A Scalable Tree Boosting System[C]// Proceedings of the 22nd ACM Sigkdd International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2016:785-794.

[15] Vaswani A, Shazeer N, Parmar N, et al. Attention Is All You Need[J]. Advances in Neural Information Processing Systems, 2017, 30:1-11.

[16] 董广. 基于异常知识发现和增量学习的银行反洗钱系统设计[D]. 杭州: 杭州电子科技大学, 2009.

[17] (Dong G. Design of Bank Anti-money Laundering System Based on Anomaly Knowledge Discovery and Incremental Learning[D]. Hangzhou: Hangzhou Dianzi University, 2009.)

[18] 沈捷. 数据挖掘技术在银行反洗钱领域中的应用研究[D]. 杭州: 浙江大学, 2007.

[19] (Shen J. Research on the Application of Data Mining Technology in Bank Anti-money Laundering[D]. Hangzhou: Zhejiang University, 2007.)

[20] 郑迎飞, 陶文纳, 赵旭, 等. 人机耦合反洗钱监测系统构建与机器学习算法优化[J]. 系统管理学报, 2021, 30(6):1198-1206.

[21] (Zheng Y F, Tao W N, Zhao X, et al. Construction of Human-Machine Coupling Anti-money Laundering Monitoring System and Optimization of Machine Learning Algorithms[J]. Journal of Systems Management, 2021, 30(6):1198-1206.)

[22] Weber M, Domeniconi G, Chen J, et al. Anti-money Laundering in Bitcoin:Experimenting with Graph Convolutional Networks for Financial Forensics[EB/OL].(2019-07-31)[2025-03-01]. https://arxiv.org/abs/1908.02591.

[23] Alarab I, Prakoonwit S, Nacer M I. Competence of Graph Convolutional Networks for Anti-money Laundering in Bitcoin Blockchain[C]// Proceedings of the 2020 5th International Conference on Machine Learning Technologies. New York: Association for Computing Machinery, 2020:23-27.

[24] Cardoso M, Saleiro P, Bizarro P. LaundroGraph:Self-supervised Graph Representation Learning for Anti-money Laundering[C]// Proceedings of the Third ACM International Conference on AI in Finance. New York: Association for Computing Machinery, 2022:130-138.

[25] Kanezashi H, Suzumura T, Liu X, et al. Ethereum Fraud Detection with Heterogeneous Graph Neural Networks[EB/OL].(2022-03-23)[2025-01-25]. https://arxiv.org/abs/2203.12363.

[26] Rao S X, Zhang S, Han Z C, et al. xFraud:Explainable Fraud Transaction Detection[EB/OL].(2020-11-24)[2025-03-13]. https://arxiv.org/abs/2011.12193.

[27] Bouritsas G, Frasca F, Zafeiriou S, et al. Improving Graph Neural Network Expressivity via Subgraph Isomorphism Counting[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(1):657-668.

[28] Chen Z D, Chen L, Villar S, et al. Can Graph Neural Networks Count Substructures?[J]. Advances in Neural Information Processing Systems, 2020, 33:10383-10395.

[29] Singh A, Jain A. An Empirical Study of AML Approach for Credit Card Fraud Detection-Financial Transactions[J]. International Journal of Computers Communications & Control, 2019, 14(6):670-690.

[30] Toyotaro S, Hiroki K. Anti-money Laundering Datasets[EB/OL].(2021-01-01)[2025-01-18]. https://github.com/IBM/AMLSim.

[31] Weber M, Chen J, Suzumura T, et al. Scalable Graph Learning for Anti-money Laundering:A First Look[EB/OL].(2018-11-30)[2025-02-07]. https://arxiv.org/abs/1812.00076.

[32] Altman E, Blanuša J, Von Niederhäusern L, et al. Realistic Synthetic Financial Transactions for Anti-money Laundering Models[J]. Advances in Neural Information Processing Systems, 2024, 36:29851-29874.

[33] Xu Y S, Li J Q, Wu A Q. An XGBoost Algorithm Based Model for Financial Risk Prediction[J]. Technical Gazette, 2024, 31(6):1898-1907.

基金资助

国家自然科学基金(62372494)

吉林省科技发展计划项目(20220201145GX)