![]()
青岛大学计算机科学技术学院
摘要: 针对现有社区检测算法中普遍存在的表示学习与聚类分离及忽视社区结构层面信息的局限性，导致社区识别效率与精度受限的问题，提出了MBCD模型。该模型通过优化图卷积网络架构，融合马尔可夫稳定性和节点间的相似度信息，以捕捉多尺度的社区结构，更精确地识别社区边界。实验表明，与多种现有的社区检测模型相比，MBCD模型在处理非重叠与重叠社区检测任务时展现出显著性能提升，证实了其在提升社区检测效能方面的有效性和优越性。关键词: 社区检测;图卷积网络;社区结构稳定性;节点相似性
基金资助:教育部人文社会科学规划基金项目（21YJA860001）； 山东省自然基金面上项目（ZR2021MG006）
![]()
A Community Detection Algorithm From A Dual-view PerspectiveLUO Kai, BIN Sheng, SUN GengxinCollege of Computer Science & Technology, Qingdao University,
Qingdao 266071, China)Abstract: In response to the
common problems in existing community detection algorithms, such as the
separation of representation learning and clustering and the neglect of
information at the community structure level, which leads to limited efficiency
and accuracy of community identification, this study proposed the MBCD model.
The model optimizes the graph convolutional network architecture and integrates
Markov stability and similarity information between nodes to capture
multi-scale community structures and identify community boundaries more
accurately. Experiments show that compared with a variety of existing community
detection models, the MBCD model exhibits significant performance improvement
when dealing with non-overlapping and overlapping community detection tasks,
which strongly confirms its effectiveness and superiority in improving the
performance of community detection.Keywords: community detection; graph convolutional networks; stability
of community structure; node similarity第一作者：骆凯（1999-），男，山东德州人，硕士研究生，主要研究方向为复杂网络。
通信作者：孙更新（1978-），男，山东青岛人，副教授，主要研究方向为复杂网络，近年来着重于探索复杂网络中传播动力学及相关传播模型。骆凯,宾晟,孙更新.一种双视角下的社区检测算法[DB/OL].[2025-01-08].http://kns.cnki.net/kcms/detail/37.1402.N.20250108.1044.002.html.
