All Times ET
Virtual
Modeling and Forecasting of the COVID-19 Outbreak in Various Indian States (304218)
*Abhishek Bhattacharjee, Dept. of Bioengineering, University of Illinois at Urbana-ChampaignVeena Mendiratta, Nokia Bell Labs
Keywords: COVID-19, modeling, forecasting, public health, India, USA
This project sought to model and forecast the COVID-19 pandemic within various states in India to see how demographically and geographically different areas are being affected. These trends were then compared to known virus behavior in the United States. Additionally the project used various models to compile forecasts for COVID-19 case counts in a given region to track and predict the spread of the disease. The project utilized the Exponential Smoothing, Auto Regressive-Integrated-Moving Average, LSTM, Susceptible-Infected-Recovered (SIR), and Prophet models to accomplish this. Data for this analysis was pulled from the covidregionaldata R package found on CRAN.
Via an analysis done through the SIR and Prophet models, rural states were found to be growing cases (proportionate to their population) faster than urban states. This trend is mostly different from the US where urban areas generally have faster moving pandemic activity, except in the case where the rural region exceeds a certain case threshold. The forecasting portion of the project found that the SIR and Prophet models were the best predictive models. These models averaged under 2% error for a 7-day forecast.