The community discovery method in heterogeneous networks: a survey

Abstract Community detection can uncover and analytically interpret a network’s topological structure, revealing its developmental patterns, internal configurations, and evolutionary processes. As a result, community detection has garnered widespread attention in the academic community. Most existing community detection research focuses on homogeneous information networks with a single node type and relationship. However, most real-world networks are heterogeneous, comprising various node types and connections. Hence, community detection algorithms for heterogeneous networks hold great application potential and research value. This article analyzes heterogeneous network community detection methods and summarizes existing approaches. First, we provide an overview of heterogeneous network community detection and define the issues in this field by examining the relevant literature. Next, we introduce the various community detection methods for heterogeneous networks and their primary evaluation metrics, categorizing existing approaches according to different network structures and algorithms. This includes community detection based on meta-paths, tensor decomposition, and deep learning approaches. Finally, we summarize and project the future trends in heterogeneous network community detection methods.