计算机科学

计算机科学 ›› 2025, Vol. 52 ›› Issue (6): 118-128.doi: 10.11896/jsjkx.240400033

• 数据库&大数据&数据科学 • 上一篇    下一篇

自适应建模网络动力学的动态链路预测方法

郭翾1, 侯锦霖1, 王文俊1, 焦鹏飞2   

  1. 1 天津大学智能与计算学部 天津 300350
    2 杭州电子科技大学网络空间安全学院 杭州 310018
  • 收稿日期:2024-04-04 修回日期:2024-09-03 出版日期:2025-06-15 发布日期:2025-06-11
  • 通讯作者: 王文俊(wjwang@tju.edu.cn)
  • 作者简介:(guoxuan@tju.edu.cn)
  • 基金资助:
    海南省重点研发项目(ZDYF2024SHFZ051);国家自然科学基金(62372146);浙江省自然科学基金(LDT23F01015F01);新奥集团合作项目(2023GKF-1220)

Dynamic Link Prediction Method for Adaptively Modeling Network Dynamics

GUO Xuan1, HOU Jinlin1, WANG Wenjun1, JIAO Pengfei2   

  1. 1 College of Intelligence and Computing,Tianjin University,Tianjin 300350,China
    2 College of Cyberspace Security,Hangzhou Dianzi University,Hangzhou 310018,China
  • Received:2024-04-04 Revised:2024-09-03 Online:2025-06-15 Published:2025-06-11
  • About author:GUO Xuan,born in 1996,postgraduate,is a member of CCF(No.U5970M).His main research interests include graph machine learning and complex network analysis.
    WANG Wenjun,born in 1970,Ph.D,professor,Ph.D supervisor,is a member of CCF(No.13864S).Hismain re-search interests include computational social sciences,big data mining and complex network analysis.
  • Supported by:
    Key R&D Program of Hainan(ZDYF2024SHFZ051),National Natural Science Foundation of China(62372146),Zhejiang Provincial Natural Science Foundation(LDT23F01015F01) and ENN Group Project(2023GKF-1220).

PDF (PC)

105

摘要： 动态网络链路预测是理解和分析动态网络的核心问题之一。针对链路预测面临的捕获复杂网络结构和真实演化规律等困难的问题,提出了一种融合图神经网络和神经常微分方程的自适应网络动力学建模方法——双层活跃度约束神经常微分方程模型DANOM。DANOM融合节点的重要性和相对位置信息,增强了网络结构的表征;通过节点活跃度约束下的神经常微分方程单元强化了演化规律的学习过程;并在节点活跃度和节点表示的重构损失优化下,挖掘到网络的有效信息。DANOM在多个真实数据集上的多种下游任务中均达到了最优效果,其中在单步链路预测任务中AUC与AP最高分别提升14%和9.7%,在快照缺失情况下的链路预测任务中AUC与AP分别平均仅降低0.43%和0.03%,在用户缝合实验中AUC与AP最高分别提升20.6%和24.4%。

关键词: 图表示学习, 动态网络, 链路预测, 神经常微分方程, 网络动力学

Abstract: Dynamic network link prediction is one of the core issues in understanding and analyzing dynamic networks.In response to the challenges of capturing complex network structures and real evolution patterns faced by link prediction,this paper proposes a method integrating the graph neural network and neural ordinary differential equation to adaptive model various network dynamics:double-layer activity-constrained neural ordinary differential equation model(DANOM).DANOM integrates the importance and relative positional information of nodes to enhance the representation of network structures,strengthens the learning process of evolution patterns through neural ordinary differential equation units constrained by node activity,and mines effective information of the network under the reconstruction loss of node activity and node representation.DANOM achieves optimal results in various down-stream tasks on multiple real-world datasets.It achieves the highest improvements of 14% and 9.7% in terms of AUC and AP,respectively,in the single-step link prediction task.In cases of snapshot missingness,the average AUC and AP of link prediction are only reduced by 0.43% and 0.03%,respectively.In the user stitching experiments,DANOM achieves the highest improvements of 20.6% and 24.4% in terms of AUC and AP,respectively.

Key words: Graph representation learning, Dynamic network, Link prediction, Neural ordinary differential equation, Network dynamics

中图分类号: 

  • TP391
[1]CAO Z,FAN Z,WANG Q,et al.Link Prediction AlgorithmBased on Denoising Autoencoder in Complex Networks[J].Journal of Chinese Computer Systems,2023,44(3):665-672.
[2]CAO Y,DONG Y H,WU S Q,et al.Advances in Dynamic Network Representation Learning Research [J].Acta Electronica Sinica,2020,48(10):2047-2059.
[3]QIN M,YEUNG D Y.Temporal Link Prediction:A UnifiedFramework,Taxonomy,and Review [J].ACM Computing Surveys,2023,56(4):1-40.
[4]ZHANG W,LAI X,WANG J.Social Link Inference via Multiview Matching Network from Spatio-Temporal Trajectories [J].IEEE Transactions on Neural Networks and Learning Systems,2020,34(4):1720-1731.
[5]PARK N,LIU F,MEHTA P,et al.Evokg:Jointly Modeling Event Time and Network Structure for Reasoning over Temporal Knowledge Graphs [C]//Proceedings of the 15th ACM International Conference on Web Search and Data Mining.2022:794-803.
[6]KING I J,HUANG H H.Euler:Detecting Network LateralMovement via Scalable Temporal Link Prediction [J].ACM Transactions on Privacy and Security,2023,26(3):1-36.
[7]QIN M,ZHANG C,BAI B,et al.High-Quality Temporal Link Prediction for Weighted Dynamic Graphs via Inductive Embedding Aggregation [J].IEEE Transactions on Knowledge and Data Engineering,2023,35(9):9378-9393.
[8]GOYAL P,CHHETRI S R,CANEDO A.Dyngraph2vec:Capturing Network Dynamics Using Dynamic Graph Representation Learning [J].Knowledge-Based Systems,2020,187:104816.
[9]MIN S,GAO Z,PENG J,et al.STGSN-A Spatial-TemporalGraph Neural Network Framework for Time-Evolving Social Networks [J].Knowledge-Based Systems,2021,214:106746.
[10]CHEN R T,RUBANOVA Y,BETTENCOURT J,et al.Neural Ordinary Differential Equations [C]//Proceedings of the 32nd International Conference on Neural Information Processing Systems.2018:6572-6583.
[11]THOMAS N K,MAX W.Semi-Supervised Classification withGraph Convolutional Networks [C]//Proceedings of the 5th International Conference on Learning Representations.2017:1-10.
[12]GOYAL P,KAMRA N,HE X,et al.DynGEM:Deep Embedding Method for Dynamic Graphs [C]//IJCAI International Workshop on Representation Learning for Graphs.2018:1-8.
[13]GAO C,ZHU J,ZHANG F,et al.A Novel RepresentationLearning for Dynamic Graphs Based on Graph Convolutional Networks [J].IEEE Transactions on Cybernetics,2022,53(6):3599-3612.
[14]PAREJA A,DOMENICONI G,CHEN J,et al.Evolvegcn:Evolving Graph Convolutional Networks for Dynamic Graphs [C]//Proceedings of the 34th AAAI Conference on Artificial Intelligence.2020:5363-5370.
[15]YOU J,DU T,LESKOVEC J.ROLAND:Graph LearningFramework for Dynamic Graphs [C]//Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.2022:2358-2366.
[16]LEI K,QIN M,BAI B,et al.GCN-GAN:A Non-Linear Temporal Link Prediction Model for Weighted Dynamic Networks [C]//IEEE INFOCOM 2019-IEEE Conference on Computer Communications.2019:388-396.
[17]YANG M,LIU J,CHEN L,et al.An Advanced Deep Generative Framework for Temporal Link Prediction in Dynamic Networks [J].IEEE Transactions on Cybernetics,2019,50(12):4946-4957.
[18]SANKAR A,WU Y,GOU L,et al.Dysat:Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks [C]//Proceedings of the 13th International Conference on Web Search and Data Mining.2020:519-527.
[19]ZANG C,WANG F.Neural Dynamics on Complex Networks[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.2020:892-902.
[20]JIN D,HEIMANN M,ROSSI R A,et al.Node2bits:Compact Time-and Attribute-Aware Node Representations for User Stitching [C]//Proceedings of the 29th Joint European Confe-rence on Machine Learning and Knowledge Discovery in Databa-ses.2019:483-506.
[21]CALVO M,MONTIJANO J I,RANDEZ L.A Fifth-Order Interpolant for the Dormand and Prince Runge-Kutta Method [J].Journal of Computational and Applied Mathematics,1990,29(1):91-100.
[22]PÓSFAI M,BARABASI A L.Network Science [M].Cam-bridge,UK:Cambridge University Press,2016:1-57.
[23]GAO J,RIBEIRO B.On the Equivalence between Temporal and Static Equivariant Graph Representations [C]//Proceedings of the 39th International Conference on Machine Learning.PMLR,2022:7052-7076.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发