Deep Learning for COVID-19 prediction

Abstract

From January 30, 2020, COVID-19 disease was announced by the World Health Organization (WHO) as a Public Health Emergency of International Concern (PHEIC). For that, many scientific researchers were interested in developing algorithms and models in order to mitigate the spread of this epidemic. Existing mathematical models including compartmental models such as SEIR, SIR, SIRQ and statistical models such as ARIMA, ARMA often fail to capture the dynamic of the propagation of an epidemic. Recently, artificial intelligence-based models have proven their effectiveness and accuracy in classification and prediction tasks. This paper aim to deploy a Recurrent Neural Network architecture called Long Short-Term Memory (LSTM) neural network for predicting the next COVID-19 recovered cases in USA, India and Italy for seven days ahead. The model' effectiveness is then evaluated on the basis of the Mean Absolute Percentage Error (MAPE) criterion. Experiments show that LSTM model is accurate with a minimal error that not exceed 3%.