Undergrads awarded by NASA, European Space Agency for project predicting COVID hotspots

The team earned the Best Use of Science award at the NASA Space Apps COVID-19 Challenge for their machine learning model that can predict hotspots for COVID outbreaks

Five undergrad students majoring in computer science and biomedical engineering were recognized at a global 48-hour hackathon hosted by NASA to tackle the many challenges surrounding the COVID-19 pandemic. The team, calling themselves Michiganders Researching Coronavirus, earned the Best Use of Science award at the NASA Space Apps COVID-19 Challenge for their machine learning model that can predict hotspots for COVID outbreaks. The team was also recognized by the European Space Agency with the Euro Data Cube Award, which will support their plans to expand the project going forward.

Team members Solomon Chang (3rd year Bio Med), Sky Chen (3rd year CS), Gabe Garfinkel (4th year CS and Philosophy, Politics and Economics), Alexander Hsia (3rd year CS), and Eric Lian (3rd year CS) used data from NASA and the Japan Aerospace Exploration Agency (JAXA) to collect information on the temperature, humidity, and light exposure of each county in the US. Compiled alongside incidence and mortality rates from COVID, the different datasets were used to produce a “hotspot index” based on a formula the team designed, and then fed to their machine learning model to look for patterns. Their method was based on the idea that the amount of light produced in a region at night is correlated with greater urbanization.

Called Project Prometheus, the team’s aim was to better balance government intervention and economic activity to minimize the negative impact to communities.

“We came together because we believe COVID is a serious health crisis and we want to do something about it while we’re all quarantined at home,” says Lian.

Image Enlarge
The team's interactive map shows the machine learning model's prediction of hotspot index, a metric the students created to describe how much of a hotspot a county is.

The team’s process involved collecting a variety of incongruous data from the agencies, typically delivered in different file formats, and standardizing it all into a single database with Python. After merging the data, it was cleaned and split into training and evaluation subsets. The machine learning model then utilized the previous two week’s worth of input to predict a day’s hot spot index a week into the future.

To submit their work, the team designed a choropleth map of the US visualizing predicted hot spots and built an interactive website to present their findings.

“Being new to Javascript and web development in general, there was a lot I had to learn and implement in a very short period of time,” Garfinkel says.

In the future, the team has several potential expansions for the project that they plan to pursue with the aid of the Euro Cube Award, including exploring more complex models, reducing the amount of missing data, and designing a more sophisticated website.

“We might also create a pipeline to automatically collect satellite and coronavirus data for each day, update the database, and improve the model’s accuracy,” they write on their project site.

The team members brought different specialities to the event, both technical and in their secondary areas of study. Alex Hsia, Sky Chen, and Eric Lian focused on data processing and machine learning, while Solomon Chang and Gabe Garfinkel focused on front-end web development.

“Although most of our team is majoring in CS, some of us study other majors such as political science, economics, and bioengineering, which bring a diverse set of skills to the table,” Lian says.

For most of the members, the COVID 19 hackathon was their first ever. The challenge virtually hosted over 15,000 participants from 150 countries, with access to several different space agencies’ earth observation data. The goal of the competition was to show how satellite information can aid in the understanding of the COVID-19 outbreak on both global and local scales. Michiganders Researching Coronavirus chose to tackle the “Human Factors” category, which directed teams to identify patterns between human activity and COVID-19 cases.