社团先验与图注意力结合的重叠社团发现算法

Overlapping Community Discovery Algorithm Combining Community Prior and Graph Attention

摘要

重叠社团发现可揭示节点在多个群体中的从属关系,在复杂网络分析中具有重要意义.针对现有图神经网络方法普遍依赖单一结构信息、在社团边界模糊情况下识别性能受限的问题,提出了一种基于社团先验的图注意力重叠社团发现算法.该算法结合图注意力机制与标签传播算法,引入社团结构先验以增强节点的结构感知能力,并通过对比学习机制强化嵌入空间中社团结构的区分性.此外,设计通道交互模块,有效整合多源异构特征,进一步提升节点表示能力.在Facebook、Computer Science和Engineering 3个真实网络数据集上的实验结果表明,该方法在重叠归一化互信息指标上显著优于多种基线方法,验证了其有效性.

Abstract

Overlapping community detection reveals the membership of nodes in multiple groups and plays a significant role in complex network analysis.Existing graph neural network methods often rely on single structural information,leading to limited performance when community boundaries are ambiguous.A graph attention network algorithm for overlapping community detection based on community priors was proposed to address this issue.The algorithm combined graph attention mechanisms with the label propagation algorithm,introducing community structure priors to enhance the structural awareness of node representations.A contrastive learning mechanism was applied to strengthen the distinction of community structures in the embedding space by pulling together nodes within the same community and pushing apart those from different communities.In addition,a multi-channel feature fusion module was designed to integrate heterogeneous features from multiple sources,further improving the quality of node representations.Experimental results on three real-world network datasets—Facebook,Computer Science,and Engineering—demonstrated that the proposed method significantly outperformed various baseline approaches in terms of overlapping normalized mutual information,which verifies its effectiveness.

关键词

重叠社团发现 / 图注意力网络 / 结构先验 / 对比学习

Key words

discovery of overlapping communities / graph attention network / structural prior / contrastive learning

[1] CHEN Y B,ZHANG Y L.Overlapping community detection algorithm using global and local mutual information of graph[J].Application Research of Computers,2023,40(8):2375-2381.
[2] LI H,CHEN F C,ZHANG J P,et al.Survey of community detection algorithms in complex network[J].Application Research of Computers,2021,38(6):1611-1618.
[3] Han Z X,Shi L L,Liu L,et al.H-louvain:hierarchical louvain-based community detection in social media data streams[J].Peer-to-Peer Networking & Applications,2024,17(4):2334-2353.
[4] Kim N,Yoon Y.Effective urban region representation learning using heterogeneous urban graph attention network(hugat)[J].IEEE Access,2025,13:102602-102612,doi:10.1109/ACCESS.2025.3577202.
[5] XU Y.Novel algorithm of overlapping community detection and analysis with improved ant colony algorithm[J].Application Research of Computers,2020,37(5):1375-1379.
[6] FortunatoS.Community detection in graphs[J].Physics Reports,2010,486(3):75-174.
[7] Barrett J,DeWolfe R,Kamiński B,et al.Improving community detection via community association strength scores[C]//Modelling and Mining Networks,2025:81-95.
[8] Su X,Xue S,Liu F Z,et al.A comprehensive survey on community detection with deep learning[J].IEEE Transactions on Neural Networks and Learning Systems,2024,35(4):4682-4702.
[9] Shchur O,Günnemann S.Overlapping community detection with graph neural networks[J].arxiv preprint arxiv:1909.12201,2019.
[10] SismanisK,Potikas P,Suliou D,et al.Overlapping community detection using graph attention networks[J].Future Generation Computer Systems,2025,163:107529,doi:10.1016/j.future.2024.107529.
[11] Gu X,Huang Q W,Yang J.Overlapping community detection in vehicular social networks based on graph attention autoencoder[J].Sensors,2025,25(8):2601,doi:10.3390/s25082601.
[12] Zhuo Z J,Chen B L,Yu S B,et al.Overlapping community detection using expansion with contraction[J].Neurocomputing,2024,565:126989,doi:10.1016/j.neucom.2023.126989.
[13] Lu M L,Luo L H,Zhang X F.Ocdib:an information bottleneck-guided approach for overlapping community detection[C]//International Joint Conference on Neural Networks(IJCNN),2024:1-8.
[14] Das S,Biswas A.Overlapping community detection based on facets of social network:an empirical analysis[C]//4th International Conference on Advances in Electrical,Computing,Communication and Sustainable Technologies(ICAECT),2024:1-6.
[15] Noor M H,Ige A O.A survey on state of the art deep learning application and challenges[J].Engineering Applications of Artificial Intelligence,2025,159:111225,doi:10.1016/j.engappai.2025.111225.
[16] Xie J R,Szymanski B K,Liu X M.Slpa:uncovering overlapping communities in social networks via a speaker-listener interaction dynamic process[C]//IEEE 11th International Conference on Data Mining Workshops,2011:344-349.
[17] Li Z Y,Wang J,Li X X,et al.Speaker diarylation with overlapping community detection using graph attention networks and label propagation algorithm[J].arxivpreprint arxiv:2506.02610,2025.
[18] Zhou Z,Hu Z,Cai S M,et al.Glscl:graph local similarity contrastive learning for recommendation[J].Expert Systems with Applications,2025,284:127855,doi:10.1016/j.eswa.2025.127855.
[19] Wu J Y,Gan W H,Lu H S,et al.Graph contrastive learning on multi-label classification for recommendations[J].ACM Transactions on Intelligent Systems and Technology,2025,16(4):19,doi:10.1145/3725854.
[20] Nouranizadeh A,Far F T,Rahmati M.Contrastive representation learning for dynamic link prediction in temporal networks[J].arxiv preprint arxiv:2408.12753,2024.
[21] Velicˇkovic′ P,Cucurull G,Casanova A,et al.Graph attention networks[J].arxiv preprint arxiv:1710.10903,2018.
[22] Li R L,Wang C,Shang R H,et al.Node classification based on structure migration and graph attention convolutional crossover network[J].Knowledge:Based Systems,2025,309:112813,doi:10.1016/j.knosys.2024.112813.
[23] Nie R J,Wang G Y,Liu Q,et al.Link prediction for attribute and structure learning based on attention mechanism[J].Applied Soft Computing,2025,179:113268,doi:10.1016/j.asoc.2025.113268.
[24] Yang X,Zhao X J,Shen Z Q.A generalizable anomaly detection method in dynamic graphs[C]//Proceedings of the AAAI Conference on Artificial Intelligence,2025:22001-22009.
[25] Li Y C,Chen J L,Chen C,et al.Contrastive deep nonnegative matrix factorization for community detection[C]//IEEE International Conference on Acoustics,Speech,and Signal Processing(ICASSP),2024:6725-6729.
[26] Mcauley J,Leskovec J.Discovering social circles in ego networks[J].ACM Transactions on Knowledge Discovery from Data,2014,8(1):28,doi:10.1145/2556612.
[27] Fan Y C,Yao Y H,C J W,Gcn-se:attention as explainability fornode classification in dynamic graphs[C]//IEEE International Conference on Data Mining(ICDM),2021:1060-1065.
[28] Yang J,Leskovec J.Overlapping community detection at scale:a nonnegative matrix factorization approach[C]//Proceedings of the 6th ACM International Conference on Web Search and Data Mining,2013:587-596.
[29] Li Y,Sha C F,Huang X,et al.Community detection in attributed graphs:an embedding approach[C]//Proceedings of the AAAI Conference on Artificial Intelligence,2018:338-345.
[30] He C B,Zheng Y L,Cheng J W,et al.Semi-supervised overlapping community detection in attributed graph withgraph convolutional autoencoder[J].Information Sciences,2022,608:1464-1479,doi:10.1016/j.ins.2022.07.036.
[31] Muttakin M N,Hossain M I,Rahman M S.Overlapping community detection using dynamic dilated aggregation in deep residual gcn[J].arxiv preprint arxiv:2210.11174,2024.
附中文参考文献:
[1] 陈燕兵,张应龙.联合图全局和局部互信息的重叠社团检测算法[J].计算机应用研究,2023,40(8):2375-2381.
[2] 李 辉,陈福才,张建朋,等.复杂网络中的社团发现算法综述[J].计算机应用研究,2021,38(6):1611-1618.
[5] 许 英.利用改进蚁群算法的重叠社团检测分析方法[J].计算机应用研究,2020,37(5):1375-1379.

基金资助

国家自然科学基金项目(61803264)资助.