The rise and fall of COVID-19 cases depends on countless factors, but new Brock University research is able to harness that data and predict various scenarios related to the virus.

Through a joint project with University of Alberta researchers, Assistant Professor Pouria Ramazi of Brock’s Department of Mathematics and Statistics is using machine learning and artificial intelligence (AI) algorithms to make these critical predictions.

Funded by Alberta Innovates, the project sees COVID-19 related data sets, such as Google mobility data, temperature and number of intensive care unit beds, fed into an AI algorithm to help predict the outcome of proposed preventive policies and to advance scientific research.

Before making a decision, scientists and policy-makers can use these algorithms to predict the outcomes of their policy decisions up to 10 weeks in the future.

“Machine learning often works through mass exposure to data,” said Ramazi, who is working alongside University of Alberta Professors Hao Wang, Russ Greiner, David Wishart and Mark Lewis on the project. “With enough data points presented over many iterations, the algorithm can almost instinctually recognize patterns and predict the outcomes.”

For example, the number of people congregating in public places combined with current mask policies and the number of people allowed in a social circle may indicate the total rise in cases in the future.

“There are endless combinations of data that machine learning can process,” said Ramazi. “If designed properly, machine learning may surpass classical research methods. It has found interesting contributing factors to COVID-19 trends, such as the total number of meat packaging plants in the area.”

Ramazi finds the algorithms useful for their ability to notice patterns in data that humans might not think to review.

Scientists must decide which of the many contributing factors to include in their research when using conventional methods.

“Machine learning says, ‘include them all and let me decide for you,'” Ramazi said.

This predictive response is similar to the way humans recognize faces.

Rarely is a person ‘reasoning’ the thousands of data points present in a human face, Ramazi said. Humans do this rapidly and with ease from a young age by being exposed to faces repeatedly over a long period of time. It becomes possible to see the pattern of a face instantaneously.

Machine learning can be said to work on the same principle.

In addition to influencing decision-making, the outcomes created by the machine can be used by other scientists to fuel their research.

Ramazi believes the research team’s machine learning prototype will continually improve its predictive accuracy.

“We find the best results within a month in the future, although we can predict further out now,” he said. “Currently, the models have a 10 per cent margin of error in predicting the future 10 weeks.”

Faculty of Mathematics and Science Dean Ejaz Ahmed understands the importance of accurate data and model predictions.

“Now more than ever, we need talented researchers like Pouria Ramazi to use these powerful new tools to predict how COVID-19 will affect us all,” he said. “Pouria’s work may contribute to important advancements needed to end the pandemic sooner.”