Proposal for heuristics-based refinement in clustering problems

Community detection in networks has recently obtained a huge interest in both natural and social sciences, for its variety of implications and applications. Several algorithms and strategies have been proposed up to now, mainly focusing on speed optimization or on the quality of the nal clustering obtained. The main scope of this paper is to bridge these two ap- proaches, via the introduction of heuristic schemes, which may be intended as a ranking of nodes inside a graph, indicating those who pose major prob- lems in community assignment. By performing the slowest, most eective algorithms (renement) on only a small fraction of the whole network, this approach is made applicable to huge networks.