Local-entity resolution for building location-based social networks by using stay points (Q2220818)

This paper studies local-entity resolution for building location-based social networks by using stay points. By using a multilevel approach, the stay points is utilized to define local-vertices that represent venues in reality. The initial network is created by using the information from stay point trajectories. Vertices are then grouped to generate super vertices until a solution network emerges to model the real world scenarios. In this approach, the coarsening process is scalable and new matching protocols can be implemented conveniently. The local-entity resolution is referred to as the task of defining local-vertices in the network. The method is divided into three phases. The initial network is built based on the set of stay point trajectories. The coarsening process is based on a matching algorithm to generate super vertices. The final network solution phase creates the networks with interactions between vertices suitable for real applications. The proposed algorithms have been tested on two empirical datasets: the data set with check-ins performed in Gowalla and Brightkite. Both of them are location-based social networks in the United States during the period from 2008 to 2010.