Persistent Community Detection Method Based on Nonuniform Network Snapshot Partitioning

Persistent community detection contributes to a deep understanding of the core components of dynamic networks. The existing persistent community detection methods have the defect of inaccurate granularity setting of network snapshot partitioning, leading to high computational overhead and inaccurate results of persistent community search. This paper proposes a persistent community detection method based on nonuniform network snapshot partitioning, which includes two steps: network snapshot partitioning and persistent community search. The nonuniform partitioning of network snapshots is carried out by the constructed community division result evaluator. In the persistent community search step, similar communities on the obtained network snapshots are filter according to the similar community pruning condition. To expand the duration of the filtered communities on the network snapshots, the community stability discriminator is used to judge whether a node subset is a community on a given network snapshot. Once the duration of the community reached the threshold, it was included in the persistent community set. Experimental results on real datasets show that the proposed method is superior to the persistent community detection method based on uniform network snapshot partitioning, and the proposed community stability discriminator can identify persistent communities more accurately.