Forecasting the progression of Covid-19 in some worst-hit Indian states using a Bayesian structural time series model

Abstract

Background: Numerous studies dealing with analysis for the future patterns of COVID-19 in different countries using conventional time series models is available [4] [5]. This study attempts to Compare the traditional time series models with Bayesian structural time series models. The study also aims to understand the causal impact of government regulations on COVID infection rates. Methods: We have used the Bayesian structural time series (BSTS) models to predict the temporal trajectory of infection spread (COVID-19) in India's worst affected states. We have also analyzed government regulations' casual impact in specific states post-Diwali using intervention analysis under BSTS models. Results: We achieved better accuracy levels in BSTS models compared to conventional time series models (ARIMA models). The forecasts for the next 10 days suggest that all states are expected to have a decreasing trend of daily cases except for Rajasthan. The prediction for Rajasthan state was an increasing pattern, possibly because of a second wave infection. The effect of local lockdowns in Rajasthan and Delhi seems to have had a significant impact in reducing the infection spread after Diwali. In the state of Gujarat, the intervention analysis suggests no statistical relation. Conclusion: On the whole, India as a country is having a decline in daily reported cases and daily deaths. Specific states like Delhi, Rajasthan, and Gujarat imposed local lockdowns to prevent another infection wave caused due to Diwali related activities. The government regulation measure seems to have a causal effect in reducing the infection spread rate.