| Paper ID | MLSP-17.6 |
| Paper Title |
GRAPH FREQUENCY ANALYSIS OF COVID-19 INCIDENCE TO IDENTIFY COUNTY-LEVEL CONTAGION PATTERNS IN THE UNITED STATES |
| Authors |
Yang Li, Gonzalo Mateos, University of Rochester, United States |
| Session | MLSP-17: Graph Neural Networks |
| Location | Gather.Town |
| Session Time: | Wednesday, 09 June, 14:00 - 14:45 |
| Presentation Time: | Wednesday, 09 June, 14:00 - 14:45 |
| Presentation |
Poster
|
| Topic |
Machine Learning for Signal Processing: [MLR-APPL] Applications of machine learning |
| IEEE Xplore Open Preview |
Click here to view in IEEE Xplore |
| Virtual Presentation |
Click here to watch in the Virtual Conference |
| Abstract |
The COVID-19 pandemic severely changed the way of life in the United States. From early scattered regional outbreaks to current country-wide spread, the contagion exhibits diverse patterns at various timescales and locations. We thus conduct a graph frequency analysis to investigate the COVID-19 spread patterns in different US counties. Commute flows between 3142 US counties were used to construct a graph capturing the population mobility. The numbers of daily confirmed COVID-19 cases per county were collected and represented as graph signals, which were then mapped into the frequency domain via the graph Fourier transform. The concept of graph frequency in Graph Signal Processing (GSP) enables the decomposition of graph signals (i.e., daily confirmed cases) into modes with smooth or rapid variations with respect to the underlying graph. These different modes of variability are shown to relate to COVID-19 spread patterns within and across counties. Changes in the nature of spread within geographical regions are also revealed by graph frequency analysis at finer temporal scales. Overall, our GSP-based approach leverages case count and mobility data to unveil spatio-temporal contagion patterns of COVID-19 incidence for each US county. Results here support the promising usage of GSP for epidemiology knowledge discovery on graphs. |
