World leaders don’t have a crystal ball to help them predict how the pandemic will look in the coming months, but two teams of UCF researchers are working to provide the next best thing.

The teams — known as University of Central Florida and Pandemic Wave Predictor —have advanced with 46 other teams to phase two of the Pandemic Response Challenge. The UCF teams, which were selected out of 104 semifinalists from 28 countries, include faculty and graduate students from the departments of Statistics and Data Science and Computer Science.

The competition gauges teams’ accuracy predicting COVID-19 infection rates and their ability to grade the efficiency of government policies like screening international travelers and school shutdowns. Teams’ submission could also guide decisions on the quantity and location of vaccine distribution — as well as land them among the winners, who will be awarded $500,000 in prizes by the host, XPRIZE. The organization is a non-profit that promotes technology development through public competitions.

The UCF teams have advanced through phase one, which started in November last year and ended earlier this month. By Feb. 3, they will complete phase two by submitting an intervention plan that may be able to assist regional governments, communities and organizations in reducing impacts from the pandemic.

Winners will be selected on how close their predictions match against real outcomes and will be announced on Feb. 26.

The University of Central Florida Team

The team named University of Central Florida is comprised of Dongdong Wang, a computer science doctoral student; Timothy Sumner, a master’s student in statistics; Zihang Zou, a computer science doctoral student; Shunpu Zhang, chair and professor in UCF’s Department of Statistics and Data Science; and Liqiang Wang, a computer science associate professor.

Predictions based on hard numbers minimizes guesswork and maximizes impact, says Zhang.
“That’s what big data analytics can solve,” Zhang says. “If you dig deeper through seemingly messy data, you can find some truths.”

The value of the research lies in distinguishing association from causation, Wang says. For instance, determining if there is there a link between a dip in cases and canceling public events or a spike in the numbers and opening up public transportation.

Last year, the University of Central Florida team used AI and deep-learning models to conduct COVID-19 case predictions months before entering the competition. The data science group has diligently continued to apply their expertise in this area and submits their numbers weekly to the Centers for Disease Control and Prevention for inclusion in government forecasts. They are currently pursuing more accurate ways to measure average data.

The Pandemic Wave Predictor Team

The Pandemic Wave Predictor team is comprised of computer science professor Lotzi Bölöni and computer science doctoral students Sharare Zehtabian and Siavash Khodadadeh.

This team’s computer model works by applying advanced machine learning to epidemiological models to learn their parameters, including when to recommend certain interventions for helping reduce the spread of COVID-19.
A unique approach of their model is that it takes into account cultural differences that can affect COVID-19 responses and preventative measures, such as the degree to which a culture is individualistic or collective.

“We believe that policy decisions should be made on an informed basis, not based on gut feeling,” Bölöni says. “Models such as ours can present the policymakers with information about the outcomes they can expect if certain kinds of interventions are mandated. It also takes into account the economic and human cost factors of the interventions, such as the closing of the schools, as well as cultural factors.”

The Pandemic Wave Predictor team is currently working on the second part the model, which is generating the intervention prescriptions.