Community identity in a temporal network: a taxonomy proposal

Abstract

Communities of nodes are one of the most important meso structures in a network. In static networks they are essentially characterized by the nodes they hold, how modular they are and how they connect to other communities. Once identified, they do not change. In networks that evolve over time, communities can shed and acquire new nodes. This generates new constructs and raises the question of community identity, and of the characterization of the events that define their lifecycle. Although researchers have devoted efforts to address some of these questions, we believe that a formalized classification and a principled method to identify community events is still lacking. In this paper we propose such a classification in the form of a robust taxonomy, supported by a similarity metric based on the Jaccard index but adjusted to chance, and a set of rules that unequivocally can track a community journey from "cradle to grave".