The Impact of Spatial Resolution and Representation on Human Mobility Predictability

Western society is distinguished by its mobility. At no time in our history have we enjoyed the capacity to travel as rapidly, conveniently and safely. On the surface, this might suggest that human mobility patterns are highly irregular and impossible to predict. Drawing on a detailed multisensory positioning data set, we replicate earlier cell tower based predictability analyses with granular spatial and temporal multisensory data, and demonstrate a spatial resolution dependence of entropy, while reinforcing the claims of inherent predictability of human mobility advanced in early works. We demonstrate that mobility entropies reported with GPS data remain essentially unchanged with pruning of noisy GPS signals, lending additional credence to our methodology. We further compare cell tower results to those from WiFi-based localization for exactly the same time periods and participants, and demonstrate that the finer spatial resolution of WiFi also results in reported entropy exceeding that from cell tower traces, indicating that the resolution dependence observed for GPS data is not entirely due to GPS noise or discretization representation. This work represents a significant step towards fundamental understanding of human mobility patterns, which serve as key mediators to policy design in fields as diverse as public health, urban planning, and delay tolerant networks.