WCSE 2023
ISBN: 978-981-18-7950-0 DOI: 10.18178/wcse.2023.06.024

HOCD: Overlapping Community Detection based on Hypergraph Convolutional Neural Networks

Dengdi Sun, You Shi, Bin Luo, Zhuanlian Ding

Abstract—Community structure is one of the universal topological attributes in complex networks. Finding community structure is the basic task of complex network analysis. Community detection is designed to divide the network into multiple substructures, which plays an important role in understanding the network and revealing the potential function of the network. As the amount of data increases, the network becomes more complex. In particular, nodes in a network may have multiple identities, such as node features and whether communities overlap. In recent years, many community detection methods based on deep learning have made great progress. However, the problem of these works is that the representation ability of models is limited. How to design an efficient and powerful representation model is very important. We have noticed that the hypergraph model can detect the higher-order relationship in the data. We use the powerful representation ability of the hypergraph convolutional network and apply it to the task of community detection, and finally realize our idea, and our method has achieved the best performance in the experiment.

Index Terms—Overlapping community detection, Hypergraph convolution neural networks

Dengdi Sun
School of Artificial Intelligence, Anhui University, CHINA
Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, CHINA
You Shi, Bin Luo
School of Computer Science and Technology, Anhui University, CHINA
Zhuanlian Ding
School of Internet, Anhui University, CHINA


Cite: Dengdi Sun, You Shi, Bin Luo, Zhuanlian Ding, "HOCD: Overlapping Community Detection based on Hypergraph Convolutional Neural Networks" Proceedings of 2023 the 13th International Workshop on Computer Science and Engineering (WCSE 2023), pp. 171-180, June 16-18, 2023.