An Artificial Intelligence (AI)-based algorithm developed by the University of Witwatersrand (Wits) has shown there is a low risk of a third wave of COVID-19 infections across South Africa in the next two weeks.
The early detection system – which was designed in partnership with iThemba Labs, the Gauteng Provincial Government, and Canada’s York University – predicts future daily confirmed cases based on historical data of previous infections in the country.
It recognises an additional wave when the actual value of daily cases deviates from the predicted value for each province in South Africa.
“The current data shows us the risk for a third infection wave of COVID-19 is small across most of provinces in South Africa, but we still remain highly vulnerable,” said Professor Bruce Mellado, director of the Institute for Collider Particle Physics at Wits University.
Mellado was confident that the model presented a very good prediction over at least a two-week period.
While predictions can be made over longer periods, they become less accurate over time.
The model has been trained on the interim period in between waves one and two in all of the South African provinces. It was tested with data taken during the period of past peaks to evaluate its performance.
How it works
The model includes features such as mobility indices, stringency indices, and epidemiological parameters.
“These parameters are consistent with clinical public health measures that can contain, control and mitigate against the COVID-19 pandemic,” said Dr. James Orbinski, Director of the York University Dahdaleh Institute for Global Health Research.
The AI-based algorithm works in parallel and supports the data of an already existing algorithm that is based on more classical analytics.
Wits said that the existence of two independent algorithms added robustness to the predictive capacity of the algorithms.
Each of the algorithms work independently and are updated on a daily basis.
Wits said the advent of COVID-19 infection waves was driven by circumstances that were difficult to predict and to control.
“In this complex environment, early detection algorithms can provide an early warning to policy makers and the population.”
“AI is very effective in navigating through complex problems with a large number of parameters and dimensions, while at the same time learning from the data,” Wits stated.
The data of the AI-based analysis is published on a website that is updated on a daily basis.
The model currently shows a higher risk of a third wave of infections in the Free State and Eastern Cape.
More populous provinces such as Gauteng, the Western Cape, and KwaZulu-Natal are currently at a very low risk of another wave of infections.
The graphs for these five provinces can be seen below.
The orange line indicates the risk threshold for the possible onset of the third wave in each province.
The red line shows the actual risk index. When this is consistently above the orange line, the probability of a wave is high.