Estimation of COVID-19 Cases with Machine Learning on Graphs

Inspiration

The Coronavirus Disease 2019 (COVID-19) brought an unprecedented sanitarian crisis around the world in 2020. Several developed countries such as Italy, Spain, and France are struggling fighting against COVID-19. The capacity of some countries to detect new cases has been overwhelmed by the exponential number of cases, leading to a poor and unreliable cases estimation. As a matter of fact, Colombia stopped testing suspected cases of COVID-19 on March 27th, 2020 because of a failure in an essential machine for the COVID-19 diagnosis.

This project models the amount of COVID-19 cases in certain places as a problem of reconstruction of time-varying graph signals. This project is rooted in the rich theory of sampling and interpolation of graph signals. The end user will be able to see an estimation of the COVID-19 cases in certain location trough the web; using past data and the information of, perhaps, more reliable surrounding localities. This will help under-financed mayors and governors to take better decisions with respect to this world crisis.

What it does

This software estimates the number coronavirus cases in certain location by modeling this problem as a reconstruction of time-varying graph signals.

How I built it

For now, the core algorithm is in Matlab. Eventually, we will make our web page using Node.JS, React, and finally Python for the machine learning algorithm.