Gaussian Regression Models for Day-Level Forecasting of COVID-19 in European Countries

Abstract

Coronavirus (COVID-19) outbreak has reached a global disease and has begun to threaten all over the world. No vaccine or drug has been developed to treat the outbreak, still. According to the data of the World Health Organization (WHO), the COVID-19 virus can be easily transmitted through respiratory droplets and contact routes. States need to take necessary measures to control the outbreak. In this study, pandemic data for eight countries in the European region are analyzed between 22 January 2020 and 30 April 2020. These countries are Belgium, Netherlands, Germany, UK, Spain, France, Italy, and Turkey. The number of cases is taken into consideration in the selection of countries. Gaussian regression analysis is preferred for the COVID-19 outbreak analysis. A comparative analysis is performed using Ard-Exponential kernel function, Rational-Quadratic kernel function, and Squared Exponential kernel functions. The quasi-Newton algorithm is used to optimize regression models. Confirmed death and recovered case data are processed in the scope of the analysis. The first 90 days of the data are used for the training of Gaussian regression models. The last nine days are evaluated for prediction. Mean Absolute Error (MAE), Median Error (ME), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE) error metrics are used for metric of performance. In general, it is seen that the Ard-Exponential kernel with Gaussian Regression (AEGR) method obtained high metric performances for all cases and countries. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.