MSN‐CDF: New community detection framework to improve routing in mobile social networks

Abstract

SummaryIn mobile social networks (MSN), portable devices of mobile users communicate with each other. The contact and intercontact times are the two most important parameters that specify the behavior of MSN. Many social features such as similarity, centrality, closeness, and distance are derived from contact and intercontact time patterns. Those parameters and features can be used in the community detection process in community‐based routing as a vector weight of network graph edges. But there is no solution for proper exploitation of them as a multicriteria optimization of community detection algorithms so that the communities become more adaptive with the MSN quiddity and the computational complexity of the algorithms to be stay constant. In this paper, we propose the MSN‐adaptive community detection framework, MSN‐CDF, which meets these demands. We improve the quality function of community detection methods, for example, the Intensity in clique percolation methods and the Modularity in modularity‐based methods, through our new model‐based joint optimization of their quality functions. To this end, to express the MSN contact and social features, we develop a criterion called MCSF. It holds the contact time and intercontact time distributions parameters and implies their social features for MSN purposes. This criterion has as little traffic as possible and enjoys from the probabilistic joint distribution model. The simulations show that MSN‐CDF reduces the network overhead significantly compared with the DiBUBB, Epidemic, and Dynamic Routing. Moreover, it outperforms DiBUBB and Dynamic Routing in terms of delivery rate and delivery latency.