Abstract
Since the onset of the Covid-19 pandemic, numerous challenges have emerged, including ensuring an adequate supply of personal protective equipment, evaluating the sufficiency of the healthcare workforce, and determining safety measures to sustain businesses and the economy. Consequently, there is a critical need for a computationally competent and realistic model to monitor current caseloads and forecast future cases, thereby enhancing public health awareness, preparation, and response. However, many forecast models currently in use have wide prediction intervals, diminishing their effectiveness as forecasting tools. Thus, this study aims to analyse the trend of Covid-19 cases in Malaysia and develop a forecast model that provides appropriate limits to improve prediction accuracy. This study relied on secondary data of daily Covid-19 cases in Malaysia provided by Ministry of Health from April 12, 2021, to April 24, 2022. Future Covid-19 incidence was predicted using simple, double and Holts-Winter exponential smoothing and SARIMAX models. SARIMAX (0, 1, 1) (1, 0, 2)7 was identified as the best model, exhibiting the lowest error values for forecasting cases. However, the results indicated that SARIMAX's prediction intervals were broad. To address this issue, a new model called hybrid SARIMAX-SARIMA was proposed where the orders from the best SARIMAX model found by using auto.arima() function are extracted and used to specify the order for a SARIMA model. The resulting combined model was then utilized to predict future trends in daily Covid-19 cases and evaluation during cultural festivals and state elections. It was observed that the proposed model outperformed others, demonstrating lower error rates and narrower confidence intervals for future predictions.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have