AN LSTM Model to predict the COVID-19 New Cases for European Countries

Abstract

Since the emergence of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), more than 510 million people have been infected, and more than 6 million deaths have been recorded globally. Accurate estimations for the number of new cases are a crucial step toward controlling and ending the pandemic. The accurate prediction helps decision-makers to prepare for future resource allocation and set the mitigation policies accordingly. The goal of this study is to develop a predictive model that can capture the patterns using the set of past policies, weather conditions, and the historic number of COVID-19 cases to predict future cases. To achieve this goal, we developed a predictive model based on long short-term memory (LSTM) and trained it using a combination of the policies implemented and the weather data to predict the number of new COVID-19 cases in European countries. The results show that the model is capable of capturing the pattern successfully and can predict future new cases. LSTM model outperformed the baseline models, including ridge regression, least absolute shrinkage selection operator (lasso), and multi-layer perceptron (MLP). The results of this study will be used as critical inputs for developing a framework that prescribes policies for the future such that it can reduce the number of new cases while keeping the implementation or post-effect costs as low as possible.