U.S. County-Level COVID-19 Case Prediction Features

Behzad Vahedi from the University of Colorado Boulder created this dataset to accompany the paper "Spatiotemporal Prediction of COVID-19 Cases using Inter- and Intra-County Proxies of Human Interactions". It contains features for predicting county-level new COVID-19 cases across the coterminous United States over 1- to 4-week horizons. The data includes spatiotemporal lags of infection rates, human interaction proxies, mobility data, and county socioeconomic composition.

Use Cases

• Forecasting COVID-19 case counts based on historical infection rate lags.
• Modeling the impact of human mobility patterns on disease spread.
• Analyzing how socioeconomic factors influence county-level pandemic trajectories.
• Developing and validating spatiotemporal autoregressive models like STXGB.

Strengths

• Data is designed for a specific, peer-reviewed spatiotemporal prediction model (STXGB).
• Features include multiple data types: infection rates, human interactions, mobility, and socioeconomic composition.
• Covers the coterminous United States at the county level.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

University of Colorado Boulder

Collection Method

Likely aggregated from public health and mobility data sources for research.

Geography

Coterminous United States (county-level)