Dynamic predictability and spatio-temporal contexts in human mobility

TL;DR

This paper investigates how spatio-temporal patterns in individual human mobility reveal predictability states that encode contextual and activity signatures, enabling more advanced short-term mobility predictions.

Contribution

It introduces the concept of predictability states in human mobility and demonstrates their potential for improving data-driven mobility forecasting methods.

Findings

Existence of contextual and activity signatures in predictability states

Predictability states carry significant information about mobility regularities

Potential for advanced, higher-order mobility prediction approaches

Abstract

Human travelling behaviours are markedly regular, to a large extent, predictable, and mostly driven by biological necessities (\eg sleeping, eating) and social constructs (\eg school schedules, synchronisation of labour). Not surprisingly, such predictability is influenced by an array of factors ranging in scale from individual (\eg preference, choices) and social (\eg household, groups) all the way to global scale (\eg mobility restrictions in a pandemic). In this work, we explore how spatio-temporal patterns in individual-level mobility, which we refer to as \emph{predictability states}, carry a large degree of information regarding the nature of the regularities in mobility. Our findings indicate the existence of contextual and activity signatures in predictability states, pointing towards the potential for more sophisticated, data-driven approaches to short-term, higher-order mobility predictions beyond frequentist/probabilistic methods.