The Effects of Weather Conditions on COVID-19 Forecasting, the United Arab Emirates as a Reliable Study Case

Abstract

Predicting new COVID-19 cases was, and still is, of paramount importance to decision-makers in many countries. Due to its transmission nature, e.g., sneezing, coughing, and physical contact, researchers have developed prediction models that include weather features hoping to improve the forecasting models' predictions. The research did not show any conclusive evidence about the importance of including weather features in forecasting models. Thus, this paper addresses this problem by considering the United Arab Emirates (UAE) COVID-19 cases and weather conditions. Using long-short term memory (LSTM) models, a variant of artificial neural network used for forecasting, we compare a uni-variate, default forecasting model that only considers COVID-19 cases to other bi- and multi-variate models that relies on COVID-19 and weather features. The results show that including weather features in the forecasting models did not significantly improve the accuracy of the default LSTM model; the maximum increase in the coefficient of determination did not exceed 0.02. Moreover, humidity, if considered with other weather features, has a small influence on improving the prediction accuracy.