Comparing Spatial Social Network Models for an Agent-based Simulation

Abstract

Location-Based Social Networks (LBSNs) combine location information with social networks and have been studied vividly in the last decade. The main research gap is the lack of available and authoritative social network datasets. Publicly available social network datasets are small and sparse, as only a small fraction of the population is captured in the dataset. For this reason, network generators are often employed to synthetically generate social networks to study LBSNs. In this work, we propose an evolving social network, implemented in an agent-based simulation, to generate realistic social networks. In the simulation, as agents move to different places of interest have the chance of making social connections with other agents as they visit the same place. The choice of places that agents visit in our simulation is informed by a large-scale real-world mobility dataset. 

We show qualitatively that our simulated social networks are more realistic than traditional social network generators including the Erdos-Renyi, Watts-Strogatz, and Barabasi-Albert.