Now Reading
Facebook Releases Model that Predicts the Evolution of the Pandemic

Facebook Releases Model that Predicts the Evolution of the Pandemic

Facebook has decided to make available in open source code the Artificial Intelligence (AI) model developed by the company and which aims to make a prediction of the evolution of the pandemic. To do this, the American giant trained the model with public data and based on different factors – such as the number of confirmed cases, symptom surveys, mobility rates, testing data, and even weather data.

Now, this model can be used by programmers, universities, companies and even public health entities. “We hope it will be used by [pandemic] response teams,” said Maximillian Nickel, a researcher at Facebook’s Artificial Intelligence research unit, at an international event with journalists in which Exame Informática participated.

According to data shared by the technology company, Facebook’s prediction model is among those with better results, for having a lower prediction error rate and for having a higher prediction quality than other of the main models currently used.

Also according to Maximillian Nickel, the Facebook model is ideal for predicting the evolution of Covid-19 in the medium term – one, two, three weeks – but not so much in the short term (e.g. in three days) or long term (e.g. in three months).

The scientist and researcher at the US technology company points out that although the company has reached a point where it has total confidence in the model to the point of making it publicly available, the path to ‘fine-tuning’ the forecasting system was not simple. In addition to inconsistent and incomplete data, Covid-19 itself evolved differently in different regions of the same country and even from country to country.

“What you can see is that there are different growth curves in different countries. One of the main goals of the project was to develop a model that addresses these challenges,” added the Facebook spokesperson.

Another situation that ‘messed’ with the effectiveness of the model was the beginning of the vaccination process, where the error rate skyrocketed. But once Facebook started integrating this variant of information into the model as well, the results improved significantly. “The model is very data-driven. We can train the system to make sense of different data and then plot a prediction,” explained Maximillian Nickel.

The AI model now released by Facebook is actually not new, being tested since March 2020 with New York University. As the pandemic has unfolded, Facebook has found new partners, including two Europeans who joined the project in October last year – the University of Vienna in Austria and the Polytechnic University of Catalonia (UPC) in Spain, the latter of which has been responsible for generating reports on the evolution of the pandemic for the European Commission.

See Also

Enric ÁLvarez-Lacalle, professor at UPC, says that the Facebook model has already been used by hospitals in the Barcelona region and attests that in the medium term, “Facebook’s Artificial Intelligence has a better predictive capacity.” “And in hospitals the concern is with next week,” he stresses.

SUBSCRIBE TO GET OUR NEWSLETTERS:

Scroll To Top

We have detected that you are using AdBlock Plus or other adblocking software which is causing you to not be able to view 360 Mozambique in its entirety.

Please add www.360mozambique.com to your adblocker’s whitelist or disable it by refreshing afterwards so you can view the site.