The UPC research group Computational Biology and Complex Systems (BIOCOM-SC) will work in collaboration with the Germans Trias i Pujol Research Institute with the support of Facebook Artificial Intelligence (Facebook IA) to train, test and calibrate new models based on artificial intelligence and developed by the company to forecast the evolution of the pandemic in the US and Austria.
The BIOCOM-SC group is working with Facebook IA to validate whether the algorithm can be adapted to make one-month forecasts, first in the south of Europe and then in the entire European continent.
On 20 October, Facebook Artificial Intelligence (Facebook AI) began to publish forecasts of the spread of COVID-19 in over 3,000 US counties. These forecasts use Facebook’s Data for Good tools, including the Symptom Survey and Movement Range Maps. They are available on the Humanitarian Data Exchange website and published on the Data for Good site.
The core of the algorithm is a neural network that is designed to predict the evolution of COVID-19 using past data. This network must learn about the evolution of the epidemic in the past and try to show how this evolution depends on patterns of mobility, climate and human behaviour. Currently, modelling human responses is extremely difficult using a mechanistic approach, that is, knowing in advance how people will react. Faced with this obstacle, artificial intelligence (AI) is considered one of the best options in empirical modelling due to its capacity to infer changes in future behaviour based on past behaviour. Details of the algorithm and how it is trained have been published in open access.
The results of the forecast made using AI for the USA have shown robust behaviour in comparison with latest generation models for three-week predictions, but detailed calibration is pending. So, the BIOCOM-SC research group, associated with the Department of Physics of the UPC will collaborate with Facebook AI to improve the model for predicting the evolution of the pandemic in Spain.
Related news: COVID-19 · Mathematical models and data management in public health emergencies