Time Series Forecasting for Coronavirus (COVID-19)

ABSTRACT

The upsurge of the novel coronavirus has spread to many countries and has been declared a pandemic by WHO. It has shaken the most powerful countries across the world like the USA, UK, and has affected economies of various countries. The coronavirus or the 2019-nCoV causes the disease that has been named COVID-19. This disease transmits by inhaling droplets that are expelled by an infected person. It has been affecting people in different ways and has been found to be threatening for the older population or people with comorbidities. It has been seen that the virus 2019-nCoV spreads faster than the two of its antecedents namely severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV). No cure or vaccine has been discovered as of now and taking precautions like staying at home are the only possible solutions. Our study analyzes the current trend of the disease in India and predicts future trends using time series forecasting. The official dataset provided by John Hopkins University through a GitHub repository has been used for the research for the time period of 22 January 2020 to 31 May 2020. The trend in cases, fatalities, and the people who have recovered until the date of 31 May 2020 has been discussed in the paper. It has been seen through the findings that the total number of cases is expected to rise to 2,15,000 by the end of May 2020 i.e. 31 May 2020 as per the AR (Autoregression) model. ARIMA (Autoregressive Integrated Moving Average) model predicts the number of cases to be 2,05,000 until the same date. Actual data has shown that the number of confirmed cases is 1,90,609 as on 31 May 2020 giving a percentage error of 7.57% and 12.85% for ARIMA and AR model respectively. Comparison between the findings of the two models has been shown later in the paper. © 2021, Springer Nature Singapore Pte Ltd.