U.S. Pandemic Prediction Using Regression and Neural Network Models

摘要

With the global outbreak of COVID-19 in 2020, it is essential for government to make aware of the trend of the pandemic. To achieve this goal, some regression and neural network models are used to predict pandemic data of the U.S. Three models -linear regression, logistic regression, and Recurrent Neural Network (RNN) - are selected for predicting cases per million people in America. Then, the effectiveness of these models is compared. These models are evaluated using Mean Squared Error (MSE). It can be concluded that while the traditional regression models, including linear and logistic regression, are much more efficient for inference, RNN predicts more accurately, with the smallest MSE being nearly 2.8. This paper gives effective guidance for American governments on how to select models to predict relevant data of the pandemic.