Activity graphs: Spatial graphs as a framework for quantifying individual mobility

Human mobility is poorly captured by existing methods which employ simple measures to quantify human mobility patterns. This paper develops spatial graph-based methods to quantify patterns of human mobility-termed activity graphs. Activity graphs are constructed with anchors representing activity locations and edges connecting anchors representing movement between anchors. We first perform a factor analysis to identify four primary dimensions of mobility that can be derived from activity graphs: quantity, extent, connectedness, and clustering. A case study with GPS tracking data from a sample of UK-based workers is then used to demonstrate how activity graphs can be applied in practice and how new dimensions of mobility captured by activity graphs may lead to new insights about mobility behaviour. We provide several promising new areas for future work where activity graphs can be further extended to address increasingly sophisticated spatial questions around individual mobility. Our analysis fits within the time-geographic framework presented by Hagerstrand, and our results highlight opportunities for continued research motivated by issues emphasized by Hagerstrand in his seminal work.

Automated summary

This paper develops spatial graph-based methods to quantify patterns of human mobility termed activity graphs. Four primary dimensions of mobility are identified: quantity, extent, connectedness, and clustering. A case study with GPS tracking data demonstrates how activity graphs can be applied in practice and how new dimensions of mobility captured by activity graphs may lead to new insights about mobility behaviour. The analysis fits within the time-geographic framework presented by Hagerstrand, and highlights opportunities for continued research motivated by issues emphasized by Hagerstrand in his seminal work.