Judy Fox PhD

Judy Fox PhD

Associate Professor of School of Data Science at the University of Virginia

Covid Prediction

Funded by NSF Expedition grant

Github

Introduction

  • Framework:
    • Combines sensitivity analysis with Google’s Temporal Fusion Transformer (TFT).
    • Explains spatio-temporal features and model-learned patterns.
  • Methodology:
    • Uses Morris Method to assess feature sensitivity.
    • Correlates Morris index with attention patterns for interpretation.
  • Key Contributions:
    • Predicts COVID-19 cases at a county level with dynamic, detailed data.
    • Modified TFT outperforms the baseline Pytorch model.
  • Dataset:
    • Covers 3142 US counties.
    • Static features: age, health disparity, industry.
    • Dynamic features: vaccination, spread, transmissible cases, social distancing.
  • Results:
    • Accurately models complex interactions.
    • Provides interpretable insights for mitigation and response planning.

Publications

Interpreting County-Level COVID-19 Infections using Transformer and Deep Learning Time Series Models Khairul Islam, Di Zhu, Yingzheng Liu, Andrej Erkelens, Nick Daniello, Aparna Marathe, Judy Fox Proceedings of the IEEE Conference on Digital Health (ICDH 2023), July 2-8, 2023, Chicago.

References

Secondary industry segments Us Census Bureau, North American Industry Classification System(naics) 2017 Paperbound, ser. North American IndustryClassification System. Claitor’s Publishing Division, 2017.