The study, published in Nature, used machine learning models to map students and their close contacts and relations at the University of Notre Dame. The researchers designed one model to predict individual risk of contracting COVID-19 and the other to predict which pairs of students were likely to be close contacts.
They found that the models helped reduce the overhead of manual tracing. They also found that the data-driven model resulted in a significantly shorter average time to test for close contacts it flagged. The researchers encouraged such methods to be used alongside manual contact tracing in a wider range of environments, including workplaces.
