IEEE Computational Intelligence Magazine - August 2022 - 54

Dominik Fischer,
Sanaz Mostaghim, and
Thomas Seidelmann
Otto von Guericke University
Magdeburg, GERMANY
Exploring Dynamic Pandemic Containment Strategies
Using Multi-Objective Optimization
Abstract
T
he SARS-CoV-2 pandemic
demonstrates the vulnerability of
societies in a globalized world.
As pathogens spread at exponential
rates, rapid development of appropriate
medical treatments and distribution of
vaccinations are major challenges.
Under these circumstances, authorities
employ non-pharmaceutical interventions
(NPIs) against the spread, which
can impact the economy strongly.
Hence, there is a need for strategies
that help minimize infection without
sacrificing the economy's wellbeing.
This study explores the inherent tradeoff
characteristics of optimal control
strategies by utilizing the well-known
SEIR (susceptible, exposed, infectious,
recovered) pandemic model with an
integrated economic compartment.
The health economy dilemma (HED)
qualifies as a multi-objective optimization
(MOO) problem and the goal is to
find strategies which are optimal
regarding concurrent infections, economic
growth, and required intensity
of employed interventions. The major
contribution of this paper is to propose
a new methodology for containment
strategy exploration using MOO. The
experiments show that the resulting
solutions can contribute towards solving
the HED by supporting the identification
of optimal strategies. Specific
characteristics of pandemics are highlighted
in this novel tri-objective optimization
approach.
Digital Object Identifier 10.1109/MCI.2022.3181347
Date of current version: 19 July 2022
that minimize infection numbers and
bi-objective approaches that include
economic costs [4]-[6].
Most approaches use predefined containment
strategies or static estimations of
those strategies. Generally, only time and
duration of the interventions are optimized
[3], [4]. Common optimization
models concentrate on few pandemic
parameters, such as manipulation of the
contact rate, while neglecting other
parameters that influence the diffusion
process. The goal of this paper is to
explore novel containment
strategies
I. Introduction
The SARS-CoV-2 pandemic highlights
the importance of effective non-pharmaceutical
interventions (NPIs) to
reduce the pathogen's spreading rate [1]
and thus prevent a health-care system
collapse and peak fatalities [2]. Since
NPIs restrict individuals' behavior, they
have a large impact on both economic
and social systems. It is a major challenge
to find an appropriate strategy that
simultaneously considers health, economy
and social aspects, which are known
to be in conflict with each other. For
instance, the best strategy regarding
infection numbers in such scenarios is to
quarantine the population, but this is
not optimal for the economy. Such
problems are usually known as multiobjective
optimization problems
(MOPs). There is a lack of effective tools
for identifying optimal control strategies
that respect the multi-objective decision-making
process [3]. Machine learning
and evolutionary algorithms offer
methods to develop such tools, among
which are single objective approaches
54 IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE | AUGUST 2022
using multi-objective optimization. This
study extends the work in [7], which
integrates the pathogen diffusion with a
predictor for relative economic health
and enables the dynamic application of
control policies. However, they studied
only two strategies, namely social distancing
and lockdown, which are fixed in
duration and strength. Here, Salgotra et
al.'s approach is improved by supporting a
wider range of policies and by enhancing
the flexibility of their usage. In addition,
the SEIR model [8], [9] is extended, too.
Another contribution of this paper
is to study the above problem as a
tri-objective optimization problem:
Minimization of infection peaks, minimization
of economic damage (as in
[7]), and minimization of the cost of
containment measures. The problem is
solved by four state-of-the-art multiobjective
optimization algorithms and
their performance is compared. The
algorithms find strategies that are sensitive
to the current state of the pandemic
and are effective even at
moderate strength, although higher
©SHUTTERSTOCK.COM/DSTUIO
http://www.SHUTTERSTOCK.COM/DSTUIO
IEEE Computational Intelligence Magazine - August 2022

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - August 2022
