An Overlapping Community Detection Method In Opportunistic Networks

Opportunistic networks (ONS) is a new pattern of networks, in which thecomplete path between the source and destination node is not required and communication is doneby using encounter chance during process of nodes mobility. And it has a significant impact onachieving future pervasive computing. In-depth researches on reality networks reveal that manyreal networks not only own features of community structure, but also they are overlapping eachother and interrelated between communities. As a base to study community structure, detectingcommunity structures in networks is very crucial to study networks functions and analyzenetworks compositions. Thus overlapping community detection is the key issue to study thenetwork structures.For the problem of overlapping community detection in the opportunistic networks, a methodwas proposed, which is called Based on link-weight local expansion community detectionalgorithm (LWLE). According to contact duration and inter contact time during nodes contactingin ONS, the strength of relation between nodes was calculated as link-weight and the networkstopology was established by Sliding Windows method. And then on the network topology, a nodewas randomly selected to label a community and expanded by using local expansion method. Forthe expansion processes enough precise in the weighted-networks, based on the local fitness andthe internal density function, a new local expansion objective function was designed to controlcommunity expansion. While expanding, the node’s belongingness values to community werefirstly calculated, and choose the biggest one as the node to be expanded. And then the objectivefunction value was calculated. If the value become larger, the node was put into the community,otherwise this local area was determined as a community. And then the node that was not labeledcommunity was selected as the initial community to expanded, until all nodes were labeled. Atlast, for the deficiency of randomly selecting initial nodes and repeated calculating during localexpansion process, a local expansion’s optimization strategy was proposed to improve it byexploiting node clustering coefficient to select initial nodes in the weighted networks. To verify LWLE algorithm performance, the ONE simulator was selected to do experimentsand simulations, and the community-based mobility model was achieved on this platform. Weanalyze the data produced by ONE, comparing the correct rate with NBDE algorithm. Thesimulation results show that LWLE algorithm could get highly correct rate of communitydetection and get the more precise and stable overlapping community structure.