Smartphone and vehicular trajectories as data sources for agent-based infection spread modelling

Abstract

Agent-based modelling (ABM) is becoming a valuable tool in understanding infection spread. In this work, two real data sets – a travel survey data set and a cellular service records data set – are used as mobility inputs in ABMs of infection spread on urban and regional scales, respectively. The data sets are novel in that they were not generated for public health purposes, but are nonetheless amenable and accessible, largely due to an emerging data culture. Data processing methods have been developed to extract trajectories of agents from the data sets, where the trajectories are then used within an ABM to simulate the spread of a close-proximity contact-based infection, such as influenza-like illnesses. Two separate models of infection spread at the scale of a city and a province were developed. Simulations were run to demonstrate infection spread within the population, as a means of exploring the applicability and contribution of the data sets to infection spread modelling. The novelty of the work is derived from the integration of disparate and non-obvious data sets, and the results – by demonstrating qualitative congruence with the results of compartmentalized mathematical models – demonstrate the robustness of the ABM approach and set a framework for future work. Our intention is to highlight the emerging opportunities within a data culture to the multidisciplinary research associated with infection spread modelling without requiring a formal background in telecommunications or intelligent transportation systems in order to better understand the use of available data, its potential, and limitations.