Unlocking social network analysis methods for studying human mobility

Planning and operations in urban spaces are strongly affected by human mobility behavior. A better understanding of individual mobility is key to improve transportation systems and to guide the allocation of public space. Previous studies have discovered statistical laws of travel distances, but the topology of movement between places has received little attention. We propose to employ network modelling methods to analyze the effect of spatial and context attributes on individual movement patterns. The perspective of mobility as a network allows to explicitly regard dyadic dependencies of sequential location visits. Here, we consider two methods developed for social networks and provide a formulation of mobility networks to justify their applicability. First, we use the Multiple Regression Quadratic Assignment Procedure to test hypotheses on the influence of location attributes on mobility behavior. Secondly, Stochastic Actor-Oriented Models are applied to model the evolution of mobility networks over time. As a proof-of-concept study, we transform data from one GNSS-based and one check-in based dataset into mobility networks and present results from both methods. We find relations that appear for a majority of samples and thus seem inherent to mobility networks. The differences between individuals and the available datasets are further quantified and discussed. We conclude that the transfer of network modeling methods is an interesting opportunity to study network-related phenomena in geographic information science.