COVID Predictions and Responsible Weather Parameters for Infections in US

Darapaneni, N.; Lahiri, S.; Thakral, K.; Ravatale, A.; Bharadwaj, D.; Paduri, A. R.

The United States has recently become the country with the most reported cases of 2019 Novel Coronavirus (COVID-19). Several research organizations, regulatory agencies, and universities have been tracking and forecasting COVID infections and fatalities at the country (US) level especially since it peaked in Spring 2020 in the US. With this work, we aim to provide an understanding of the COVID-19 infections in the US. We have mainly focused to provide insights into the following areas, a.) Forecast of daily infections for the US for the period Nov 2020-Jan 2021, b.) Identify the top 3 States which are going to lead the infection rate for the said period, c.) Evaluate if weather parameters are directly responsible for infections. We have primarily used time series forecasting methods and related models like Prophet and ARIMA to determine the prediction for infections for Nov-Jan 2021. Also, various linear regression and ensemble models were analyzed to determine the features responsible for infections.We also selected 10 States based on their ranking on infection count to evaluate our selected model i.e. ARIMA for rate of infections projection of top 3 States. Overall, we have based our work on the already defined models and methods and have attempted to determine results by fine-tuning data, feature engineering, and selection techniques. © 2021 IEEE.