Analysis of Window Sizes in Prediction of Daily COVID-19 Cases using Machine Learning Models

The first case in Turkey of the COVID-19 pandemic was reported on 11 March 2020, with the number of confirmed cases rapidly rising to over 60.000 by April 2021. In the absence of effective treatment, an important tool of outbreak management is the modeling and predicting of the COVID-19 pandemic’s future and behavior. The present study considered machine learning (ML) models to predict the daily confirmed cases of the COVID-19 outbreak in Turkey. The daily confirmed cases of COVID-19 data from November 25, 2020, to June 13, 2021, were obtained from the website of the Republic of Turkey, Ministry of Health. The predicted values were explored with a range of window sizes from 2 to 14 and four ML models: Support Vector Machine (SVM), Random Forest (RF), k-Nearest Neighbor (k-NN), eXtreme Gradient Boosting (XGBoost). The results show that for the test dataset, the performance of window size 6 is better than the other window sizes for the SVM and k-NN models. For the RF and XGBoost models, the performance of window sizes 8 and 10 is better than the other window sizes, respectively.

Analysis of Window Sizes in Prediction of Daily COVID-19 Cases using Machine Learning Models Keywords:

COVID-19 , SVM , RF , k-NN , XGBoost , Turkey