An Exponential Smoothing Holt-Winters Based-Approach for Estimating Extreme Values of Covid-19 Cases

Abstract

Covid-19 is an ongoing outbreak across the world infecting millions, having significant fatality rate, and triggering economic disruption on a large scale. The demand of healthcare facility has been significantly affected by the increased Covid-19 cases. Many countries have been forced to do lockdown and physical distancing to avoid a crucial peak of novel Covid-19 pandemic that potentially overwhelms healthcare services. Central Java is the province with the third highest population density in Indonesia and predicted to be affected significantly over a particular period of this outbreak. Our paper aims to provide a modelling to estimate extreme values of daily Covid-19 cases in Central Java, between March and April 2021. We particularly capture seasonality during this period using Exponential Smoothing Holt-Winters. We employ that Value at Risk and mean excess function based-approaches for extreme value estimation. Our simulation studies indicate that Exponential Smoothing Holt-Winters and Value at Risk provide sharp and well prediction for extreme value with zero violation. Since a number of positive cases has resulted unprecedented volatility, estimating the extreme value of daily Covid-19 cases become a crucial matter to support maintain essential health services.