Predictive Modeling for Coronavirus Pandemic: A Time Series Analysis

Abstract

The COVID-19 pandemic has led to a dramatic loss of human life worldwide. India, with a population of more than 1.34 billion - the second largest population in the world have faced acute difficulty in controlling the transmission of Coronavirus among its population, particularly during the second wave. This results into serious repercussions on mortality and morbidity in India. In this Study, the secondary data which is available on public domain of Government of India and Other Countries was used. This data is extracted from different websites from 1 April, 2020 to 30 April, 2021 for a total of 395 days. This data consists of – cumulative confirmed cases, active cases, recovered cases, the actual deaths per day and cumulative deaths. The association between daily confirmed cases and mortality was established using the generalized additive model (GAM) with natural and penalized spline smoothers at (6,2,2) degrees of freedom in R software with mortality as a dependent variable. Smoothers for day of the week, active cases, actual active cases were also included in the model. In the corresponding period mortality rate on an average 533 is deaths per day. The association between daily confirmed cases and daily mortality was found to be statistically significant. The relative risk has been computed for increase in every 1000 number of daily confirmed cases. For increase in 1000 number of daily confirmed cases, the expected number of deaths amounts to 1.006702 (approximately). The entire data has been divided into seven zones and for each zone GAM Model was fitted to make future prediction. The analysis box plots, smoothing plots along with PACF
plots, residual plots and predicted plots.