A spatio-temporal network model to represent and analyze LBSNs

Abstract

With the increasing popularity of Location-based Social Networks (LBSNs), users have shared information about places they have visited, creating a link between the real world (their movements on the globe) and the virtual world (what they express about these movements on the LBSNs). In this article, we propose the SiST model, which contains information captured from different dimensions (Social, Spatial and Temporal). The proposed model is a graph that links two users, as long as both of them are friends and have published that they were at the same place within a predefined time interval. In addition to movement patterns that can be extracted using SiST, this model may be used to predict if two users will meet in a short time span by executing a classification algorithm. Performance tests were conducted with SiST networks that were built based on three real LBSN datasets. Results indicated that it is possible to forecast with high accuracy (ranging from 80.50% to 96.32%) whether two people will meet or not using two days of historical data.