ARIMA forecasting of COVID-19 incidence in Italy, Russia, and the USA

TL;DR

This study uses ARIMA models to forecast COVID-19 incidence in Italy, Russia, and the USA, providing insights into peak timings and the impact of lockdown measures on the epidemic's progression.

Contribution

It applies ARIMA modeling to COVID-19 case data across three countries, identifying optimal models and analyzing epidemic peaks and recovery timelines.

Findings

Italy peaked in mid-April, Russia in late May, USA in mid-May

ARIMA models are reliable when cases stabilize

Stricter lockdowns in Italy may have influenced epidemic dynamics

Abstract

The novel Coronavirus disease (COVID-19) is a severe respiratory infection that officially occurred in Wuhan, China, in December 2019. In late February, the disease began to spread quickly across the world, causing serious health, social, and economic emergencies. This paper aims to forecast the short to medium-term incidence of COVID-19 epidemic through the medium of an autoregressive integrated moving average (ARIMA) model, applied to Italy, Russia, and the USA The analysis is carried out on the number of new daily confirmed COVID-19 cases, collected by Worldometer website. The best ARIMA models are Italy (4,2,4), Russia (1,2,1), and the USA (6,2,3). The results show that: i) ARIMA models are reliable enough when new daily cases begin to stabilize; ii) Italy, the USA, and Russia reached the peak of COVID-19 infections in mid-April, mid-May, and late May, respectively; and iii) Russia and the USA will require much more time than Italy to drop COVID-19 cases near zero. This may suggest the importance of the application of quick and effective lockdown measures, which have been relatively stricter in Italy. Therefore, even if the results should be interpreted with caution, ARIMA models seem to be a good tool that can help the health authorities to monitor the diffusion of the outbreak.