RESEARCH: Modeling the Dynamics of the COVID-19 Population in Australia: A Probabilistic Analysis
Share
Ali Eshragh, Data Scientist, together with his co-authors Saed Alizamir Peter Howley Elizabeth Stojanovski, recently published research on the dynamics of COVID-19 on Australia’s Population.
The novel Corona Virus COVID-19 arrived on Australian shores around 25 January 2020. This paper presents a novel method of dynamically modeling and forecasting the COVID-19 pandemic in Australia with a high degree of accuracy and in a timely manner using limited data; a valuable resource that can be used to guide government decision-making on societal restrictions on a daily and/or weekly basis. The “partially-observable stochastic process” used in this study predicts not only the future actual values with extremely low error but also the percentage of unobserved COVID-19 cases in the population. The model can further assist policymakers to assess the effectiveness of several possible alternative scenarios in their decision-making processes.
You can find the highlights of the research below and or Download PDF to read the entire research paper.
Highlights
- This work applies a novel and effective approach using a partially-observable stochastic process
to study the dynamics of the COVID-19 population in Australia over the 1 March–22 May 2020
period. - The key contributions of this work include (but are not limited to):
(i) identifying two structural break points in the numbers of new cases coinciding with where the
dynamics of the COVID-19 population are altered: the first, a major break point, on 27 March
2020, is one week after implementing the “lockdown restrictions”, and the second minor point
on 18 April 2020, is one week after the “Easter break”;
(ii) forecasting the future daily numbers of new cases up to 28 days in advance with extremely low
mean absolute percentage errors (MAPEs) using a relative paucity of data, namely, MAPE of
1.53% using 20 days of data to predict the number of new cases for the following 6 days, MAPE
of 0.43% using 34 days of data to predict the number of new cases for the following 14 days,
and MAPE of 0.20% using 55 days of data to predict the number of new cases for the following
28 days;
(iii) estimating approximately 33% of COVID-19 cases as unobserved by 26 March 2020, reducing
to less than 5% after implementing the Government’s constructive restrictions;
(iv) predicting that the growth rate, prior to the Government’s implementation of restrictions, was
on a trajectory to infect numbers equal to Australia’s entire population by 24 April 2020;
(v) estimating the dynamics of the growth rate of the COVID-19 population to slow down to a
rate of 0.820 after the first breakpoint, with a slight rise to 0.979 after the second breakpoint;
(vi) Advocating the outlined stochastic model as practically beneficial for policymakers when
considering implementation and easing of virus restrictions due to the demonstrated sensitivity
of the dynamics of the COVID-19 population in Australia to both major and minor system
changes. - The model developed in this work may further assist policymakers to consider the impact of
several potential scenarios in their decision-making processes.
I hold a Bachelor’s degree (2001) and Master’s degree (2004) both in Industrial Engineering, majoring in Statistical Modeling and Stochastic Optimisation, and a PhD (2011) in Operations Research (OR) majoring in Stochastic OR. Most of the analysis I used in my PhD research was based on thorough understanding of statistical learning and Markov decision processes. The application of those results arises in Biology, Physics, Cryptography, Machine Learning and many other areas.
In 2011, I was appointed as a Postdoctoral Research Associate at The University of Adelaide in Australia. I worked on an ARC-funded research project in which I developed new stochastic models to find an optimal experimental design of a continuous-time Markovian population model. Those results have gained in popularity for modeling a range of phenomena, including Evolutionary, Ecological and Epidemiological processes.
In 2013, after successfully completing my postdoctoral research project, I was appointed as a Lecturer in Stochastic Operations Research in the School of Mathematical Sciences, the University of Adelaide. In 2014, I was offered a continuing lecturing position in the School of Mathematical and Physical Sciences, the University of Newcastle (UoN) in Australia. In 2017, I was promoted to Senior Lecturer. Currently, I am a Senior Lecturer (equivalent to Tenured Associate Professor in the U.S. system) in Statistics and Optimization at UoN.
I publish regularly in leading Operations Research/Applied Probability/Statistics journals and actively
collaborate with several outstanding national and international researchers. I am currently the Leader
of “Data Science Research Group” at UoN, an Associate Investigator in ARC Centre of Excellence for
Mathematical and Statistical Frontiers (ACEMS), Australia, and an Affiliated Researcher in the International Computer Science Institute (ICSI) at the University of California, Berkeley, U.S.A.
I have been involved in several governmental and industry funded research projects, with the total amount of AU$2,350,618. Furthermore, I have been consulting many industries spanning from service to manufacturing companies since 2000. More recently, I have been conducting several demand forecasting projects for major food and beverage supply chains in Australia.
