Forecasting of COVID-19 confirmed cases in different countries with ARIMA models
Abstract
The epidemic of a new coronavirus disease (COVID-19) has emerged as a global threat. Many countries and their health care systems were caught off guard. This study aims to predict the prevalence of COVID-19 in the most infected countries in the World Health Organization (WHO) regions in order to have better preparedness in health systems. The Auto-Regressive Integrated Moving Average (ARIMA) model was used to predict the pattern of confirmed cases based on epidemiological data from Johns Hopkins from February 25 to July 19, 2020. Mean incremental and logarithmic transfers were carried out to stabilize the series. Based on the ACF (AutoCorrelation Function) and PACF (Partial AutoCorrelation Function) charts, the first parameters of the model have been identified. The best model was chosen based on the likelihood ratio test and the least performance criteria value among all ARIMA models. Stata software version 12 was used. A number of ARIMA models have been formulated with various parameters. ARIMA (6,2,1) for South Africa, ARIMA (6,2,2) for U.S.A, ARIMA (2,1,1) for Iran, ARIMA (2,1,1) for Russia, ARIMA (5,2,2) for India, and ARIMA (3,1,2) for Australia were chosen based on the likelihood ratio tests and the values of the lower performance criteria. This research demonstrates that ARIMA models are sufficiently effective in predicting the prevalence of COVID-19 in the future. Predicting trends in COVID-19 prevalence in these countries can convince other countries to use this model in their future studies. The analysis results can help governments and health systems understand the patterns of this pandemic and plan for future waves of patients.