Forecasting COVID-19 impact in India using pandemic waves Nonlinear Growth Models

Pavan Kumar; Ram Kumar Singh; Chintan Nanda; Himangshu Kalita; Shashikanta Patairiya; Yagya Datt Sharma; Meenu Rani; Akshaya Srikanth Bhagavathula

For our study, after the pre-processing method of data smoothening and testing the database for stationary and further for prediction modeling. Therefore, a multivariate database model for COVID-19 with different interaction methods was applied. We put the model with double differencing and as per lags for observed incidence (ARIMA(1,2,0)), mortality (ARIMA(0,2,2), and recover case (Brown's method). ACF correlation is found more suited for the database, and therefore, the model we made for prediction of death, confirm, and recover variable separately is ARIMA (0,2,2) [14] . This ACF plot is understated with recovery cases and plotted in Figure 4 .