Using Rule-Based Topology Strategies to Fast Identify Complex Network Community Structures

The goal of community detection is to identify groups of nodes or network partitions with tight intra-community and loose inter-community edge connectivity. Proposed community detection approaches have made use of evolutionary computation, data mining, modularity optimization, and density- and topology-based methodologies. Here we describe our proposal for a hierarchical arcmerging algorithm that utilizes network topology and rule-based arc-merging strategies to efficiently identify community structures and network hierarchies. Five real-world social networks and eight Lancichinetti-Fortunato-Radicchi with ground-truth communities were used to verify community structure. Eight large/very large real-world networks were used to test performance efficiency, and two synthetic networks were used to determine resolution limitation problem avoidance. Our results indicate that the communities identified by the proposed method (a) were closer to ground-truth communities, (b) overcame the resolution limitation problem, and (c) revealed network hierarchies.

Project ID:PB10703-1481
External Project ID:MOST106-2221-E182-073