Schulich Data Scientists Formulate Predictive COVID-19 Dashboard
In less than a week, Schulich’s data scientists, Murat Kristal, PhD, Ikjyot Singh Kohli, PhD and David Elsner have created the COVID-19 Dynamics dashboard to provide insights into the spread of the global virus pandemic. Using publicly available data from Johns Hopkins Center for Systems Science and Engineering GIS and GitHub, the dashboard is able to predict the number of new COVID-19 cases over the next five days in each country, region, province or state.
The predictions are based on two models. The first model uses Mathematica and relies on ten days of data to predict the next 5 days. The second model uses the data science virtual machines in Microsoft Azure to predict the number of cases per day. The overall accuracy of predictions for the five-day period is illustrated in the second page of the dashboard.
“These numbers could help decision makers make informed decisions,” shared Kristal. “Even though our five-day accuracy is currently 67%, the one-day accuracy is over 90%. Given how fast things are changing this will be very helpful to them. We also expect the five-day accuracy rate to improve as more data becomes available. The important thing is the order of magnitude and our predictions provide that.”
The data is manually updated every morning, with a one-day lag. The date of the latest update is noted on the first and second page of the dashboard. In the coming days, the dashboard will have automated updates.