Abstract

The COVID-19 pandemic has been one of the most devastating events in recent history, resulting in millions of deaths and destruction to the global economy. While vaccines have been developed to slow the spread of the virus and achieve global herd immunity, millions of US citizens refuse to take them due to speculation regarding their safety and effectiveness. Our project goal is to reassure people that COVID vaccines are effective at controlling the spread of the virus and reducing the number of people infected. To demonstrate this, we use an autoregressive (AR) model and long short-term memory (LSTM) network to represent the spread of COVID over time. Using data from various US states, we display COVID trends over the last year and make predictions on how the disease will spread in the future (beyond the scope of our data set) with and without vaccines. In the end, our predictions show that vaccines are effective at reducing cases and slowing the spread of the disease. By comparing results from both models for each state, we were able to choose the more accurate model and use it for our graphs and predictions. After comparing sources of error in our models (root mean square error and coefficient of determination), our results indicated that the LSTM neural network was much more accurate than the autoregressive model.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call