Basu, Campbell's Machine Learning Model Predicts COVID-19 Spread and Mitigation Strategies

10/6/2020 Aaron Seidlitz, Illinois CS

Illinois CS PhD student Sayantani Basu paired with Sohaib and Sara Abbasi Professor Emeritus Roy H. Campbell on a machine learning model to help track and predict COVID-19 spread and effects of mitigation strategies.

Written by Aaron Seidlitz, Illinois CS

As a young teenager, Illinois CS PhD student Sayantani Basu began to code and contemplate the power found within computing. When she finished high school, a growing curiosity in healthcare led her to realize that applying computer science to medicine could benefit people all over the world.

SAyantani Basu
Sayantani Basu

Then, as she finished her final year in the Illinois CS master’s program, the COVID-19 pandemic occurred.

Basu realized the significance of the moment—computing could aid the healthcare industry in a time of need. She then began pouring her efforts into a thesis guided by her adviser, Sohaib and Sara Abbasi Professor Emeritus Roy H. Campbell.

This led to a paper – titled “Going by the Numbers: Learning and modeling COVID-19 disease dynamics" – recently published in the journal Chaos, Solitons & Fractals.

“The thing that is most interesting about disease is that it truly doesn’t discriminate,” Basu said. “It is something that everyone suffers from, no matter what place in the world you’re from, what age you are, what time in history we review. The COVID-19 pandemic is another example of this, and that notion inspired me to study it further.

“Our paper focused on tracking the disease and the mitigation measures. It feels great to have it published, so others can learn more about COVID-19 and the effects of reopening.”

The result, highlighted in the published paper, is a “Long Short-Term Memory (LSTM) based model trained on cumulative COVID-19 cases and deaths.”

Users can adjust it based on various parameters to provide predictions as needed. Results can cast a wide net and focus on the entire country or it can drill down to the county level. Insights include trends in the rate of infections and deaths and comparison of various mitigation measures.

The authors hope the paper and model can help others decide upon mitigation and reopening strategies.

“We have our open-source code set up, meaning that people are free to use it. Our primary goal is for people to integrate it with their systems to help understand the pandemic and their options to confront it,” Basu said. “A bigger goal would be for people to use this model in the future, too, during other outbreaks we know could arise."

Roy Campbell
Roy H. Campbell

Campbell said their time working together dates to Basu taking his graduate level course on health data analytics.

He noticed then the effort she put into her class project. Later, Campbell came away delighted from a conversation to find out she wanted to continue for her PhD in Computer Science.

As Basu worked toward her MS thesis, there were some hurdles and realizations along the way.

First, she had to wait out the COVID-19 data. Basu said that all machine learning projects take a significant amount of data, but as the world responded to the outbreak massive amounts of data came pouring in.

After surveying the data, she understood the way other researchers interpreted the data.

“One of the aspects she noticed is that there were not too many learning models applied to the actual data,” Campbell said. “Most of the statistical approaches leaned on simulations and models of how scientists thought people and the virus behaved. We believed, instead, that we could take the daily numbers released – of infections, deaths and more. We could learn from that data without needing to generate a behavioral model.

“It was our hope that it might offer a dynamic view of how the virus was impacting society, even as society reacted to the virus.”

It’s this kind of effort that, to Campbell, embodies the power of Illinois CS research. Young researchers, like Basu, push boundaries and produce results that benefit their specific research field.

That's what appealed to Basu, as well, considering her passion for computing and healthcare.  Even as the COVID-19 pandemic influenced the final moments of her MS program, she didn’t let anything deter her from her efforts.

“My experience at Illinois CS is something I’m so thankful for, simply because of the opportunities presented to me,” Basu said. “Of course, it’s a little sad that our in-person activities came to a halt because of the pandemic in March. But we have been working offline to compensate for the lack of in-person contact.

“I’m so appreciative of that effort, that kind of adaptability, because I can continue to pursue my academic goals.”

And it is through goals like Basu’s - who is beginning her PhD pursuit at Illinois CS - that further compels Campbell to believe students here will alter what's possible in computing.

“Computer science has a huge impact on the way we perceive the world, allowing us to produce digital models of analogue behavior,” he said. “The COVID-19 outbreak has been interesting because it was an unknown disease and human societies reacted to it in ways that were a little unpredictable.

“We proved helpful in identifying the dynamics of an unknown disease, through computer science, even as it spreads in a world full of diversity.”


Share this story

This story was published October 6, 2020.