Machine Learning Prediction Models of COVID-19 Prevalence in Turkey…


Güner E.

2021 1st International Conference On Informatics and Computer Science (ICI-CS2021), Ankara, Turkey, 9 - 11 December 2021, pp.8-14

  • Publication Type: Conference Paper / Full Text
  • City: Ankara
  • Country: Turkey
  • Page Numbers: pp.8-14

Abstract

Abstract— In December 2019, cases of pneumonia were detected in Wuhan, China. As a result of the examinations, it was reported that the disease was caused by a new coronavirus called SARS-COV-2. The disease called COVID-19 turned into an epidemic that affected the whole world in a short time. The COVID-19 pandemic caused a total of 197.905,518 cases and 4.218,403 deaths worldwide by August 1, 2021. The rapid spread of the COVID-19 epidemic has brought many serious problems with it. While countries were affected socially and economically, the biggest negative impact of the epidemic was on health systems. For this reason, it is vital to estimate the burden that will occur in a country's health system in such global epidemic situations. In this study, the prevalence of COVID-19 in Turkey was examined and the number of future cases and deaths were estimated with Linear Regression (LR), Random Forest Regression (RFR) and Bayesian Ridge Regression (BRR). The proposed estimation models were compared in terms of Coefficient of Determination (R2 ), Mean Absolute Error (MAE), Mean Square Error (MSE) and Mean Squared Deviation (RMSE). The R 2 value was found to be 1 for RFR 0.97 for BRR and 0.97 for LR. The smallest MAE, MSE, RMSE values are respectively 955,17; 1527893,16; 1236,08 belongs to the RFR model. The results showed that the RFR prediction model is more effective than other methods in estimating the spread of the epidemic in Turkey.

Keywords— Turkey, COVID-19, prediction, linear regression, random forest regression, bayesian ridge regression