Multiplicative Seasonal ARIMA Modeling and Forecasting of COVID_19 Daily Deaths in Hungary

Abstract

The coronavirus disease (COVID-19) is a terrifying pandemic that is rapidly spreading over the world. Up to this point, Hungary has had a significant COVID-19 death rate. The main purpose of this article is to model and forecast basic seasonal time series for COVID-19 death rates. The COVID 19 data, which was collected between 2020-10-04 and 2021-05-12 by the Hungarian government and the World Health Organization (WHO), has been used. The data was analyzed and models were fitted using R software version 4.1.2. The statistical time series model is fitted with the Multiplicative Seasonal Autoregressive Integrated Moving Average (SARIMA) model. Forecasts are made using the fitted model. The data output is used to find seasonality, trend patterns, and unstable variance patterns in the time series plot. The trend is made stationary using the starting difference of the converted data approach, and the variance is made constant using the logarithmic transformation of the original data set. Based on the Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) plot data, the ARIMA (1, 1, 2) (1, 0, 1) (7) model is proposed. The standardized residuals, ACF of residuals, normal Q-Q plot, and p-value for Ljung-Box statistics of the fitted model were found to be within confidence limits and to have no distinct behavioral pattern. The ARIMA (1, 1, 2) (1, 0, 1) (7) model has the smallest estimated value, with a sigma square estimated value of 0.02764, log-likelihood = 80.41, and an Akaike Information Criterion (AIC) value of 148.82. As a consequence, the fitted model ARIMA (1,1,2) (1,0,1) (7) is identified as the best model for forecasting the COVID-19 daily death rate in the country.