Time Series Forecasting of US COVID-19 Transmission.

Ding, Yongmei; Huang, Rui; Shao, Ningyi

Context The increasing number of confirmed cases of COVID-19 globally is shocking every day. US daily deaths have numbered over one-thousand people per day for nearly 3 days (from November 18, 2020 to November 20, 2020), and total deaths have exceeded 250 000 as of November 21, 2020, which drives the medical community to search for trends to provide an early warning of rising numbers of cases and to prevent future increases. Objective The study intended to evaluate available US COVID-19 data to determine the possibility of predicting the spread of COVID-19 in the USA. Design The research team collected US COVID-19 data from a time-series view and established a seasonal autoregressive integrated moving average (SARIMA) model to predict trends. Results According to the spatial and temporal distribution of cumulative confirmed cases, US COVID-19 cases are mainly concentrated in areas with high population density, with that variable having a positive correlation to the number of confirmed cases and deaths. The correlation coefficients are 0.95 and 0.817, respectively, indicating that the transmission of COVID-19 in the USA is characterized by agglomeration. After exploring the impact of population density, the research team established a SARIMA model to predict the trends, finding that US COVID-19 cases will continue to go up. Conclusions By combining knowledge of the statistical features of the virus with modeling findings, the study determined a method that can improve understanding of the serious pandemic, paving the way toward the development of predictive and preventative solutions.