Modelling COVID-19 cases and deaths with climate variables using statistical and data science methods

The authors aimed to forecast the cumulative COVID-19 confirmed cases and deaths by the most appropriate model of the top five impacted countries and three South Asian countries incorporating the climate factors as covariates. Different statistical and data science methods are used in this study. The results of the model selection criteria depict that the ELM algorithm is adequate for France, Germany, and Spain, while the MLP algorithm tends to have a better forecast for India and Pakistan. In Sri Lanka, Italy, and the United States, the well-known ARIMAX model tends to be a good match for death predictions. The findings showed that the inclusion of the meteorological variables improves the accuracy of modeling both COVID-19 cases and deaths, for all the chosen countries’ cumulative confirmed cases except Italy and Sri Lanka. However, in the case of modeling deaths, it is observed that the inclusion of meteorological variables was not able to enhance the forecasting accuracy of the model in each of the selected countries. Though no single model was found to be suitable for all countries, the authors identify the most appropriate models for each country for forecasting and make the sixty-day-ahead forecast for each country.