Task and Time Aware Community Detection in Dynamically Evolving Social Networks

The temporal analysis of the community structure in dynamically evolving networks requires that the nodes and connections between them be sampled into a time series of successive networks by shifting capturing intervals of typically equal width in time. The size of such time windows affects the outcome of community detection in certain ways possibly depending also on the detection method. In this paper we propose a systematic approach to identify time window sizes so that community detection methods produce meaningful results. For that purpose we investigate several simple indicators, which can help to sample an evolving network depending on the analysis task and the community detection method. (C) 2013 The Authors. Published by Elsevier B.V. Selection and peer review under responsibility of the organizers of the 2013 International Conference on Computational Science