Forecasting Confirmed Cases and Mortalities of COVID-19 in the US

Background The wide spread of COVID-19 in the US has placed the country as the most infected population worldwide. This paper aims to forecast the number of confirmed cases and mortalities from 12 April to 21 May, 2020. There has been a large body of literature in forecasting epidemic outbreaks such as C algorithms with shortfall of predicting for long periods and autoregressive integrated moving average models with the limited flexibility. However, the US COVID-19 data shows great variety in the relative increments of confirmed cases. This requires a reproductive time series. Method This paper suggests a time series based on the relative increments of confirmed cases. The proposed time series assumes the changes in the time series and provides flexibility. The suggested model was applied on the data observed from 27 February to 11 April 2020 and its objective is forecasting 40 days from 12 April to 21 May 2020. Results It is expected that by May 21, 2020, the accumulative number of confirmed cases of COVID-19 in the US rises to 1,464,729, with 80% confidence interval. Our analysis also shows that by the 21 st of May, the cumulative number of mortalities caused by COVID-19 in the US from 18747 cases on 11 April increases to around 73250 cases on 21 May, 2020. Conclusion Our results highlight the value of reproductive strategies in time series modelling of COVID-19. Our model benefits from a reproductive strategy from a time point in which the US COVID-19 data demonstrates a sudden fall.