Linear Regression and Holt's Winter Algorithm in Forecasting Daily Coronavirus Disease 2019 Cases in Malaysia: Preliminary Study

ABSTRACT

Coronavirus disease 2019 is a fatal viral disease presently sweeping the globe. COVID-19 is a novel coronavirus that produces an infectious illness. Thus, COVID-19 detection in the general population may be helpful. The involvement of machine learning in combating COVID-19 had rapidly increased because of its efficiency to scale up, faster processing capacity, and more dependable than humans in some healthcare activities. This paper will focus on two models which are Linear Regression (LR) model and Holt's Winter model. The COVID-19 dataset was taken from the Ministry of Health for Malaysia's website. Daily confirmed cases were recorded from 24th of January 2020 to 31st July 2021 and stored in Microsoft Excel. Waikato Environment for Knowledge Analysis (WEKA) software was utilized to perform the prediction of daily cases in the next 14-days and the quality of forecasting models is evaluated by two performance metrics, Mean Absolute Deviation (MAD) and Mean Absolute Percentage Error (MAPE). The best model is selected by the lowest value of performance metrics. The comparison shows that the forecasting result of Holt's Winter is more accurate than the LR model. The developed prediction model can help public health officials make better decisions and manage resources to decrease COVID-19 pandemic morbidity and mortality. Therefore, preparation and control procedures can be established. © 2021 IEEE.