Using Cohesive Subgroups for Analyzing the Evolution of the Friend View Mobile Social Network Alvin Chin 1 , Hao Wang 1 1 Nokia Research Center Building 2, No. 5 Donghuan Zhonglu, Economic and Technological Development Area Beijing, China, 100176 {alvin.chin, ext-hao.10.wang}@nokia.com Abstract. The mobility of users and the ubiquity of the mobile phone and Internet are leading to the development of mobile social networks. Much work has been done on modeling the evolution of online social networks using mathematical, social network analysis, and graph theoretic methods, however few using cohesive subgroups and similarity. In this paper, we present a study of the evolution of the Nokia Friend View mobile social network using network and usage statistics, and use the DISSECT method [7] for characterizing this evolution through the movement of cohesive subgroups. We discover that the friend network becomes less dense and less clustered (with fewer subgroups) over time, and the DISSECT method [7] helped to identify these cohesive subgroups and accurately predicted its most active users. We visualized these cohesive subgroups and modeled the evolution using persistence of subgroups. These results point the way towards an analytical framework for comparing mobile social networks which may help facilitate development of new recommender applications. Keywords: Mobile social network, social network evolution, social network analysis, subgroup identification, cohesive subgroups, centrality, similarity modeling. 1 Introduction Online social networks such as Facebook and LinkedIn are increasingly being used for sharing content and keeping in contact with friends, colleagues and family. With the ubiquity of the mobile phone and wireless technologies, location can be added as a context to localize the content such as location of photos taken and status updates, thus creating mobile social networks. Mobile social networking applications such as Foursquare and BuddyCloud use location of users to provide services such as finding people and places nearby, providing relevant content, providing search and updated points of interest, and creating specific topic channels from which other people can subscribe to. Previous work has studied the structure and properties of online social networks [9, 17, 18, 22] and their evolution [1, 3, 17, 22, 24]. Most use social network properties over time or create models of evolution using group formation, clustering and partitioning, or mathematical modeling and graph theory, but fail to enumerate the