Mining emerging user-centered network structures in location-based social networks

The digitization of social networks has enabled the passive collection of large scale data, which in turn have fostered social studies that have been traditionally dependent on small scale, interview-based data. During the last years, a new class of digital social networks has emerged, namely, location-based social networks (LBSNs). The main interaction between users of an LBSN is location sharing, i.e., declaring their presence to specific places. The latter ties the virtual, online world with the real space that users interact in. Thus, except from the social graph, a number of implicit network structures emerge. As an example, two people can be considered to be connected if they have been to at least k common places. Similar structures play crucial role in fields such as epidemiology and urban planning, while they can have implications in communication networks as well (e.g., mobile peer-to-peer content delivery). In this study, we examine the characteristics and the evolution of these structures using two LBSN datasets. As our analysis indicate, (i) these structures can deviate significantly from the pure social network and, (ii) they are highly dynamic (i.e., these implicit connections are ephemeral).