A stochastic approach for extracting community-based backbones

Large-scale dense networks are very parvasive in various fields such as communication, social analytics, architecture, bio-metrics, etc. Thus, the need to build a compact version of the networks allowing their analysis is a matter of great importance. One of the main solutions to reduce the size of the network while maintaining its characteristics is backbone extraction techniques. Two types of methods are distinguished in the literature: similar nodes are gathered and merged in coarse-graining techniques to compress the network, while filter-based methods discard edges and nodes according to some statistical properties. In this paper, we propose a filtering-based approach that is based on the community structure of the network. The so-called "Acquaintance-Overlapping Backbone (AOB)" is a stochastic method which selects overlapping nodes and the most connected nodes of the network. Experimental results show that the AOB is more effective in preserving relevant information as compared to some alternative methods.