Modeling the scaling properties of human mobility

TL;DR

This paper critically evaluates continuous time random walk models for human mobility using empirical mobile phone data, introduces two principles to better model trajectories, and develops a self-consistent microscopic model that predicts observed scaling laws.

Contribution

It identifies limitations of CTRW models for human mobility and proposes a new principled model that aligns with empirical data and scaling laws.

Findings

CTRW models conflict with empirical mobility data

Two principles enable a self-consistent microscopic model

Model predicts key scaling exponents accurately

Abstract

While the fat tailed jump size and the waiting time distributions characterizing individual human trajectories strongly suggest the relevance of the continuous time random walk (CTRW) models of human mobility, no one seriously believes that human traces are truly random. Given the importance of human mobility, from epidemic modeling to traffic prediction and urban planning, we need quantitative models that can account for the statistical characteristics of individual human trajectories. Here we use empirical data on human mobility, captured by mobile phone traces, to show that the predictions of the CTRW models are in systematic conflict with the empirical results. We introduce two principles that govern human trajectories, allowing us to build a statistically self-consistent microscopic model for individual human mobility. The model not only accounts for the empirically observed scaling laws but also allows us to analytically predict most of the pertinent scaling exponents.