LBSNSim: Analyzing and Modeling Location-based Social Networks

The soaring adoption of location-based social networks (LBSNs) makes it possible to analyze human socio-spatial behaviors based on large-scale realistic data, which is important to both the research community and the design of new location-based social applications. However, performing direct measurements on LBSNs is impractical, because of the security mechanisms of existing LBSNs, and high time and resource costs. The problem is exacerbated by the scarcity of available LBSN datasets, which is mainly due to the privacy concerns and the hardness of distributing large-volume data. As a result, only a very few number of LBSN datasets are publicly released. In this paper, we extract and study the universal statistical features of three LBSN datasets, and propose LBSNSim, a trace-driven model for generating synthetic LBSN datasets capturing the properties of the original datasets. Our evaluation shows that LBSNSim provides an accurate representation of target LBSNs.