Forecasting COVID-19 cases based on mobility

Abstract

Countries struggling to overcome the profound and devastating effects of COVID-19 have started taking steps to return to "new normal." Any accurate forecasting can help countries and decision-makers to make plans and decisions in the process of returning normal life. In this regard, it is needless to mention the criticality and importance of accurate forecasting. In this study, daily cases of COVID-19 are estimated based on mobility data, considering the proven human-to-human transmission factor. The data of seven countries, namely Brazil, France, Germany, Italy, Spain, the United Kingdom (UK), and the United States of America (USA) are used to train and test the models. These countries represent around 57% of the total cases in the whole world. In this context, various machine learning algorithms are implemented to obtain accurate predictions. Unlike most studies, the predicted case numbers are evaluated against the actual values to reveal the real performance of the methods and determine the most effective methods. The results indicated that it is unlikely to propose the same algorithm for forecasting COVID-19 cases for all countries. Also, mobility data can be enough the predict the COVID-19 cases in the USA.