Community discovery based on social relations and temporal-spatial topics in LBSNs

Abstract

Community discovery is a comprehensive problem associating with sociology and computer science. The recent surge of Location-Based Social Networks (LBSNs) brings new challenges to this problem as there is no definite community structure in LBSNs. This paper tackles the multidimensional community discovery in LBSNs based on user check-in characteristics. Communities discovered in this paper satisfy two requirements: frequent user interaction and consistent temporal-spatial pattern. Firstly, based on a new definition of dynamic user interaction, two types of check-ins in LBSNs are distinguished. Secondly, a novel community discovery model called SRTST is conceived to describe the generative process of different types of check-ins. Thirdly, the Gibbs Sampling algorithm is derived for the model parameter estimation. In the end, empirical experiments on real-world LBSN datasets are designed to validate the performance of the proposed model. Experimental results show that SRTST model can discover multidimensional communities and it outperforms the state-of-the-art methods on various evaluation metrics.