Author: Ahmed Alruwaili
Alruwaili, Ahmed, 2022 Stochastic process models for short-term forecasting of pandemics, Flinders University, College of Science and Engineering
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
The coronavirus disease of 2019 (COVID-19) affected the whole world economically, socially, and healthily. By the fourth quarter of 2022, over 600 million people were infected, with a mortality rate of around 6.5 million. The global response to COVID-19 management has been multi-faceted, involving restrictions, lockdowns, and immunisation programs. Forecasting models have also been widely utilised to estimate future case numbers and inform government policy. Reliable forecasts of disease case numbers are also very important from a medical perspective, as they can significantly assist with resource allocation and planning.
The scientific literature reports on an extensive range of models that have been applied to the modelling and forecasting of the COVID-19 case number dataset. Models investigated range from agent-based computational approaches to statistical stochastic process models and machine learning approaches. Many of the models investigated were successfully applied; however, many were applied in limited country-specific contexts, and substantial limitations were identified regarding the reliability of forecasts. A wealth of data on the COVID-19 pandemic has now been collected, and this data provides an opportunity to address model limitations and develop improved models for future pandemic management.
This project aims to address the gaps identified in the literature by evaluating a wide range of relevant statistical stochastic process models and neural network COVID-19 forecasting models across multiple countries. Furthermore, this project introduces and evaluates a novel modelling approach (ARMA-ELM) that combines both statistical and machine learning models. Model performance was assessed across multiple models and countries, with the ARMA-ELM providing enhanced performance in certain circumstances. Overall, significant differences were found in the COVID-19 data structures between different countries, resulting in no particular model performing best in all circumstances.
Keywords: Stochastic process models, Neural network models, COVID-19 pandemic, Forecasting
Subject: Mathematics thesis
Thesis type: Masters
Completed: 2022
School: College of Science and Engineering
Supervisor: Dr Greg Falzon