Yiming Fang, Heng Chang, Lanlan Qi* 

College of Electronic Engineering, National University of Defense Technology  

Abstract：

Community detection plays a pivotal role in complex network clustering, aiming to identify tightly- connected node subsets with dense internal links and sparse inter-community connections. Widespread in social, biological, and information networks, accurate community structure detection facilitates in-depth understanding of network functions, organizational patterns, and information propagation laws. With the increasing scale and complexity of network data, traditional community detection algorithms face significant challenges, while emerging techniques based on graph neural networks (GNNs), deep non-negative matrix factorization (DNMF), and their combinations have introduced novel research directions. This review systematically examines the state- of-the-art in traditional algorithms, GNN-based methods, DNMF-based approaches, and GNN-enhanced DNMF algorithms for community detection.

Key Words：

community detection; complex networks; graph neural networks; deep non-negative matrix factorization; modularity; spectral clustering; label propagation