A Time-Series prediction model using long-short term memory networks for prediction of Covid – 19 data

Abstract

The recent outbreak of Coronavirus disease 2019 (COVID-19), which gets caused by severe acute respiratory syndrome (SARS) coronavirus 2 (SARS-CoV-2), has been responsible for the deaths of over 3,00,000 people and at the same time has infected over 4.7 million people in the whole world as of mid-May, 2020. There has been more that 1.8 million recoveries during this period too. It becomes imperative for Governments to be aware of the situation and to be able to predict the future number of patients so that readiness in terms of health care and planning of other necessary actions can be maintained. Using the same as a strong motivation, a model for prediction of the number of COVID–19 patients has been developed using the Long-Sort Term Memory (LSTM) network and then employed it for forecasting future cases. The cases of four countries, namely, United States of America, India, Argentina and Brazil are taken into account. The study finds that the LSTM network developed in this paper performs better than two other methods known as Convolutional Neural Network and Nonlinear Auto Regressive Time Series networks and thus can be a useful candidate for prediction of future number of patients of COVID–19.