Fears loom over the U.S. due to the COVID-19 outbreak. People fear being infected by the disease. However, coronavirus test-kits are short in supply and it may usually take a few weeks for people to land a doctor's appointment. Yet,
What it does
Our WebApp provides online COVID-19 diagnosis in response to the growing pandemic. It gathers geographical, clinical, and personnel contact history data from users and estimates the user's risk of COVID-19 infection. The diagnosis also provides an action plan to users according to their risk rating.
How we built it
We built our classifier model with a decision matrix with its entries weighted by Bayesian probability estimates. We referenced expert knowledge published in China CDC's Novel Coronavirus Pneumonia Diagnosis and Treatment Plan (Provisional 7th Edition), US CDC's guidelines, and WHO's diagnosis handbook. This WebApp was built with the React.JS library in the frontend and the Flask web framework for the backend. We also connected our code to a geofencing API provided by Radar.io to tell our clients when they are about to enter an infected area.
Challenges we ran into
Statistics of the patient data have been published yet no dataset has been released to date. Our calculation is reliable statistically but still varies by case.
Accomplishments that we're proud of
We utilized Radar.io and its geofencing functions to estimate the geographical risk factor. It is proven to be a viable and convenient option. We have also put a lot of effort into enhancing user-experience since this WebApp is meant to provide trustworthy and reliable information to its users without provoking excessive fear.
What we learned
We've familiarized ourselves with Radar.io APIs and learned so much about COVID-19. We also went technical with the diagnosis model and learned a lot with probabilistic estimation.
We look forward to implementing a guide to contact hospitals based on user location. However, we do not currently have access to the Places API in Radar.io which hindered this implementation. We will also keep the classifier model up-to-date as new data is published on a daily basis.