Exact Forecasting for COVID-19 Data: Case Study for Turkey

Abstract

The novel coronavirus COVID-19 (SARS-CoV-2) with the first clinical case emerged in the city of Wuhan in China in December 2019. Then it has spread to the entire world in very short time and turned into a global problem, namely, it has rapidly become a pandemic. Within this context, many studies have attempted to predict the consequences of the pandemic in certain countries. Nevertheless, these studies have focused on some parameters such as reproductive number, recovery rate and mortality rate when performing forecasting. This study aims to forecast COVID-19 data in Turkey with use of a new technique which is a combination of classical exponential smoothing and moving average. There is no need for reproductive number, recovery rate and mortality rate computation in this proposed technique. Simulations are carried out for the number of daily cases, active cases (those are cases with no symptoms), daily tests, recovering patients, patients in the intensive care unit, daily intubated patients, and deaths forecasting and results are tested on Mean Absolute Percentage Error (MAPE) criterion. It is shown that this technique captured the system dynamic behavior in Turkey and made exact predictions with the use of real time dataset.