Evaluation of Time Series Models for Forecasting Daily Rise in Confirmed COVID-19 Cases During the Second Wave in India

ABSTRACT

The spread of Coronavirus began in Wuhan, China and very steadily went on to be a global pandemic. This virus, later identified as COVID-19 virus, was found to be extremely contagious in nature. The people contracting the virus displayed several symptoms, and in some cases, the contraction of the virus also proved to be fatal. The earliest COVID-19 case in India was found on 30th January 2020, and thereafter, the country witnessed a steady growth in the number of infections. During the following year, in the latter half of March 2021 onwards, the cases began rising exponentially indicating the start of the second wave. The intention behind this research is to predict the future of the daily COVID-19 confirmed cases in India during the second wave. This is done by utilizing three time series models, ‘Autoregressive Integrated Moving Average’ (ARIMA), ‘Long Short-Term Memory (LSTM) Neural Network’ and ‘Facebook Prophet’. The primary focus of this work is to compare and evaluate the three models to determine which model shows the least error. The results show that the performance of ARIMA is better than LSTM and Prophet. The error metrics show least amount of average error for ARIMA, followed by LSTM and Prophet. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.