COVID-19 Transmission in Connecticut

Faculty Mentor: Caiwen Ding (UConn)


COVID-19 is one of the worst pandemics in modern history. Since the first case found in China in December 2019, COVID-19 has caused over 70 million cases and 1.5 million deaths. Using past case counts to predict new case counts is an important and challenging problem for this pandemic. In this project, we propose a deep learning model based on a temporal graph convolutional network to make case count prediction at the county level in Connecticut.

A comprehensive data set of 85-days data from 8 counties in Connecticut and 10 neighbored counties are utilized. A collection of factors that might be important in predicting cases are considered, such as mobility, weather, demographics, and geographics. Our model achieves the best prediction performance when only case data is used as temporal information and distances between counties are considered in spatial information.