Project Details
The Council of State and Territorial Epidemiologists (CSTE)
Influenza-related hospitalizations are a major contributor to the overall burden of influenza in the United States. Accurate forecasts of hospital admissions can play a vital role in supporting public health planning and response, especially as multiple respiratory pathogens continue to circulate. To advance forecasting capabilities and facilitate accurate predictions, the U.S. Centers for Disease Control and Prevention (CDC) has coordinated the multi-team FluSight Challenge since the 2013–14 influenza season. Our team has participated in this effort from its inception. UVA-BI is one of the CSTE-funded sites contributing weekly forecasts of influenza-related hospitalizations and experimental categorical targets since the 2021-2022 season.
NIFFLERS: Novel Influenza Forecasting Framework using Learned Ensembles and Related Signals
Our team has developed a suite of modeling frameworks trained and updated in real time to generate weekly forecasts at both the state and national levels. The primary approach is an ensemble model that integrates outputs from multiple statistical and machine learning models (UVAFluX-Ensemble). We also employ additional spatiotemporal (UVAFluX-CESGCN) and trained ensemble (UVAFluX-OptimWISE) models that serve as secondary systems within our FluSight Challenge toolkit.
In 2024-25, this project was known as PHOENIX: PHase-Optimized Ensemble and Network approaches for Influenza forecasting and eXplanations.
FINDINGS: These models have consistently delivered robust performance across multiple influenza seasons, including high rankings during key periods of the influenza season.
OptimWISE: An Ensemble Approach for Probabilistic Forecasts Optimized via Weighted Interval Score. Submitted to 2026 CSTE Annual Conference.
Leveraging Lead-Lag Behaviors for Improved Spatiotemporal Forecasts. Submitted to 2026 CSTE Annual Conference.
Lopez et al. Challenges of COVID-19 Case Forecasting in the US, 2020–2021. PLoS Computational Biology. 2024 May 6;20(5):e1011200.
Mandal et al. OptimWISE: An Ensemble Approach for Probabilistic Forecasts Optimized via Weighted Interval Score. MIDAS Annual Meeting, 2024.
Chou et al. State-Level Spatio-Temporal Forecasting of COVID-19 Hospitalization with Uncertainty. International Symposium on Forecasting, 2023.
Adiga A, Kaur G, Wang L, Hurt B, Porebski P, Venkatramanan S, Lewis B, Marathe MV. Phase-informed Bayesian ensemble models improve performance of COVID-19 forecasts. Proceedings of the AAAI Conference on Artificial Intelligence. 2023 Sep 6; 37(13):15647-15653.
Adiga A, Kaur G, Hurt B, Wang L, Porebski P, Venkatramanan S, Lewis B, Marathe M. Enhancing COVID-19 ensemble forecasting model performance using auxiliary data sources. Proceedings of the IEEE International Conference on Big Data. 2022 Dec 17:1594-1603.
Adiga A, Wang L, Hurt B, Peddireddy A, Porebski P, Venkatramanan S, Lewis B, Marathe M. All models are useful: Bayesian ensembling for robust high resolution COVID-19 forecasting. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 2021 Aug 14:2505-2513.