Unsupervised learning for county-level typological classification for COVID-19 research

Abstract

The analysis of county-level COVID-19 pandemic data faces computational and analytic challenges, particularly when considering the heterogeneity of data sources with variation in geographic, demographic, and socioeconomic factors between counties. This study presents a method to join relevant data from different sources to investigate underlying typological effects and disparities across typologies. Both consistencies within and variations between urban and non-urban counties are demonstrated. When different county types were stratified by age group distribution, this method identifies significant community mobility differences occurring before, during, and after the shutdown. Counties with a larger proportion of young adults (age 20–24) have higher baseline mobility and had the least mobility reduction during the lockdown.