Forecasting the number of intensive care beds occupied by COVID-19 patients through the use of Recurrent Neural Networks, mobility habits and epidemic spread data

Abstract

Since 2019, the diffusion of COVID-19 all over the world has caused more than five millions deaths and the biggest economic disaster of last decades. A better prediction of the Intensive Care beds (ICUs) burden due to COVID-19 may optimize the public spending and beds occupancy, in the future. This can enable Public Institutions to apply control policies and a better regularization of regional mobility. In this work, we address the challenge of producing fully automated covid spread forecasting via Deep Learning algorithms. We developed our system by means of LSTM and Bidirectional LSTM models and new model regularization achievements such as 'Inference Dropout'. Results highlight 'state-of-Art' accuracy in terms of ICUs prediction. We definitely believe that this breakthrough can become a valuable tool for policy makers in order to face with the problem of COVID-19 effects in the near future.