Bayesian optimisation of restriction zones for bluetongue control

24^th October 2018, 13:00
Thomas Spooner
University of Liverpool

Abstract

We investigate the restriction of animal movements as a method to control the spread of bluetongue (BT), an infectious disease of livestock. Using a simulation model of BT in conjunction with Bayesian optimisation, we derive region-dependent policies that minimise the number of infected farms during an outbreak. We demonstrate that control of BT as effective as that obtained using standard UK movement restrictions can be achieved with smaller and simpler restriction zones affecting far fewer farms. It is then shown that this approach leads to, approximately, an order of magnitude decrease in the cost to farms. Our policies are also shown to be robust to perturbations in the defining variables, introduction date of the disease and spatial features of each region. Due to the increasing temperatures associated with climate change, conventional ways of tackling these problems will soon become economically infeasible. We thus conclude by recommending Bayesian optimisation as a useful technique for the design of cost-effective epidemiological control methods.

Joint work with Matthew Baylis, John Fearnley, Anne Jones, Rahul Savani, and Joanne Turner.