A Mathematical Model and Forecasting of COVID-19 Outbreak in India

Abstract

The outbreak of novel coronavirus (COVID-19) initially affected Wuhan and then rapidly spread to many nations across the globe, causing a major public health issue. In India, about 4.37 million confirmed cases and 73,890 confirmed deaths due to coronavirus were reported as of September 8, 2020. Lack of vaccination, absence of social distancing, and difficulty in full contact tracing lead to an increase of confirmed cases. We propose a dynamic model to describe the evolution of the COVID-19 pandemic and a prediction model to forecast the coronavirus epidemic. The proposed model is a modified SEIRD dynamic model with vital parameters. In addition, our model demonstrates that a reduction in the COVID-19 pandemic is possible through implementing social distancing and other suitable measures. For implementing a predication model, deep learning models are used to make predictions about the number of confirmed cases for all states across India. A recurrent neural network (RNN)-based long short-term memory (LSTM), Prophet, and autoregressive integrated moving average (ARIMA) models are used for forecasting. The accuracy obtained for the LSTM model is 74% and the error rate is calculated for Prophet and the ARIMA model. The analysis provided by our model provides valuable insights into the dynamics of spreading which is useful in curbing the epidemic.