Community Partitioning Based on Coupling Density With Asymmetric Similarity Between Vertexes

Abstract

Community partitioning has become an important approach for uncovering underlying patterns in complex network analysis. This paper proposes a coupling density model with asymmetric similarity between vertices, which considers weighted coupling relations between connected groups and vertex similarities to guide the generation of new partitions and to measure inter-community similarity. Directed similarity, as a promising quality criterion, captures the directional tendency from one vertex to another by incorporating local attributes such as degree, neighbors, co-neighbors, and their interrelations. Theoretical and empirical evaluations on both real-world and synthetic networks demonstrate that the proposed metric model outperforms traditional modularity and standard density measures. Furthermore, the model was effectively applied to analyze textile-related trade networks.