- AIpert
- Auquan Data Science
- Berkeley Yu Group
- Bob Pagano
- Carnegie Mellon University
- Center for Disease Dynamics, Economics & Policy
- Columbia University
- Columbia University and University of North Carolina
- Covid-19 Simulator Consortium
- Covid Act Now
- Discrete Dynamical Systems
- Facebook AI Research
- Federal Reserve Bank of San Francisco/Wilson
- Georgia Institute of Technology, Center for Health and Humanitarian Systems
- Georgia Institute of Technology, College of Computing
- Google and Harvard School of Public Health
- The Hong Kong University of Science and Technology
- IEM
- Imperial College, London
- Institute for Business Forecasting
- Institute of Health Metrics and Evaluation
- Iowa State University
- IQVIA Analytics Center of Excellence
- John Burant
- Johns Hopkins University, Applied Physics Lab
- Johns Hopkins University, Applied Physics Lab
- Johns Hopkins University, Applied Physics Lab, Testing and Diagnostic Working Group
- Johns Hopkins University, Center for Systems Science and Engineering
- Johns Hopkins University, Infectious Disease Dynamics Lab
- Johns Hopkins University, the University of North Carolina, and Google
- Karlen Working Group
- Lehigh University Computational Uncertainty Lab
- LockNQuay
- Loïc Pottier
- London School of Hygiene and Tropical Medicine
- Los Alamos National Laboratory
- Los Alamos National Laboratory and Northern Arizona University
- Masaryk University
- Massachusetts Institute of Technology, Cassandra
- Massachusetts Institute of Technology, COVID-19 Policy Alliance
- Massachusetts Institute of Technology, Institute for Data, Systems, and Society
- Massachusetts Institute of Technology, Laboratory for Computational Physiology
- Massachusetts Institute of Technology, Operations Research Center
- Microsoft AI
- Northeastern
- Notre Dame University
- Oliver Wyman
- OneQuietNight
- Pandemic Central
- Predictive Science Inc.
- Predictive Science Inc.
- Qi-Jun Hong
- Rensselaer Polytechnic Institute and University of Washington
- Robert Walraven
- SignatureScience
- State University of New York, Upstate Medical University & Syracuse University
- Steve Hortman
- Steve McConnell
- Texas Tech University
- University of Arizona
- University of California, Los Angeles
- University of California, Merced
- University of California, San Diego and Northeastern University
- University of California, Santa Barbara
- University of Central Florida
- University of Chicago
- University of Colorado Boulder
- University of Georgia, Center for the Ecology of Infectious Disease
- University of Geneva / Swiss Data Science Center
- University of Massachusetts, Amherst
- University of Michigan
- University of Southern California
- University of Texas, Austin
- US Army Engineering Research and Development Center
- Wadhwani AI
- Walmart Labs Data Science Team
- Youyang Gu (COVID-Projections)
Model name: AIpert
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Piecewise log linear regression model
Forecasts submitted: Deaths
Model name: Auquan
Intervention assumptions: These projections do not make specific assumptions about which interventions have been implemented or will remain in place.
Methods: Fitted SEIR model
Forecasts submitted: Deaths
Model name: Yu_Group
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Ensemble of combined linear and exponential predictors (CLEP)
Forecasts submitted: Cases
Model name: BPagano
Intervention assumptions: These projections assume that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: SIR model
Forecasts submitted: Deaths
Model name: CMU
Intervention Assumptions: These projections do not make specific assumptions about which interventions have been implemented or will remain in place.
Methods: Autoregressive time-series model
Forecasts submitted: Cases, deaths
Model name: CDDEP
Intervention Assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Bayesian SEIR model
Forecasts submitted: Cases
Model name: Columbia
Intervention assumptions: This model assumes that contact rates will increase 5% during the first week of the forecast period. Following week 1, the reproductive number is then set to 1.0.
Hospitalization assumptions: The model uses state-specific hospitalization data, when available. In states without hospitalization data, the model uses the national average value for hospitalization data.
Methods: Metapopulation SEIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: Columbia-UNC
Intervention assumptions: This model assumes that transmission intensity will peak in early July and then gradually decline.
Methods: Statistical survival-convolutional model
Forecasts submitted: Cases, deaths
Model name: Covid19Sim
Intervention assumptions: This model is based on assumptions about how levels of social distancing will change in the future.
Hospitalization assumptions: The number of new hospitalizations per day are estimated from the number of infections, using state-specific hospitalization rates.
Methods: SEIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: CAN
Intervention assumptions: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
Methods: Fitted SEIR model
Forecasts submitted: Hospitalizations, deaths
Model name: DDS
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Bayesian hierarchical model
Forecasts submitted: Cases, deaths
Model name: Facebook
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: A machine learning model, combined with an auto-regressive model
Forecasts submitted: Cases
Model name: FRBSF-Wilson
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: An SIR-derived econometric county panel data model with transmission rate assumed to be function of weather and mobility
Forecasts submitted: Cases
Model name: GT-CHHS
Intervention Assumptions: This model assumes that that once stay-at-home orders are lifted, contact rates will gradually increase. It also assumes that some households containing symptomatic cases will self-quarantine.
Methods: Agent-based model
Forecasts submitted: Deaths
Model name: GT-DeepCOVID (formerly: GA_Tech)
Intervention Assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Hospitalization Assumptions: Daily hospitalizations are predicted from publicly available, state-level data sources.
Methods: Deep learning
Forecasts submitted: Hospitalizations, deaths
Model name: Google-HSPH
Intervention assumptions: These forecasts implement changes to future population mobility in order to predict COVID-19 transmission intensity.
Methods: SEIR model fit with machine learning
Forecasts submitted: Cases, hospitalizations, deaths
Model name: HKUST-DNN
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Deep neural networks trained on observed deaths, cases, and hospitalizations to forecast state-level cumulative deaths
Forecasts submitted: Deaths
Model name: IEM
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: SEIR model with machine learning.
Forecasts submitted: Cases
Model name: Imperial
Intervention Assumptions: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
Methods: Ensembles of mechanistic transmission models, fit to different parameter assumptions
Forecasts submitted: Deaths
Model name: IBF
Intervention Assumptions: These projections do not make specific assumptions about which interventions have been implemented or will remain in place.
Methods: Combination of a mechanistic disease transmission model and a curve-fitting approach
Forecasts submitted: Cases, deaths
Model name: IHME
Intervention Assumptions: Projections are adjusted to reflect differences in aggregate population mobility and community mitigation policies.
Hospitalization assumptions: Daily hospitalizations are estimated from predictions of daily deaths, using state hospitalization rates, where available.
Methods: Combination of a mechanistic disease transmission model and a curve-fitting approach
Forecasts submitted: Cases, hospitalizations, deaths
Model name: ISU
Intervention Assumptions: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
Methods: Nonparametric spatiotemporal model
Forecasts submitted: Cases, deaths
Model name: IQVIA
Intervention Assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Spatiotemporal attention network model
Forecasts submitted: Cases
Model name: JCB
Intervention Assumptions: The incidence of COVID-19 in the population determines the strength of and resulting impact of control measures in the future.
Methods: Phenomenological statistical model
Forecasts submitted: Deaths
Model name: JHU-APL
Intervention Assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Metapopulation SEIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: JHUAPL-Gecko
Intervention Assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: SARIMA time series model (1,1,0)x(1,1,0,7) with anomaly detector applied to confirmed hospital admissions since 09/01/2020.
Forecasts submitted: Hospitalizations
Model name: JHUAPLTDWG-ICATTML
Intervention Assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Ensemble of models forecasts provided with lagged data derived from the ICATT program, CELR tests, and wastewater surveillance.
Forecasts submitted: Hospitalizations
Model name: JHU-CSSE
Intervention Assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Machine learning
Forecasts submitted: Cases, deaths
Model name: JHU-IDD
Intervention Assumptions: These projections assume that current interventions will not change during the forecasted period.
Hospitalization Assumptions: Daily hospitalizations are estimated from predictions of daily cases. A standard proportion is applied to all states.
Methods: Metapopulation SEIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: JHU-UNC-Google
Intervention Assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: This forecast is an ensemble of two different models: A multiplicative growth model and a curve fitting model.
Forecasts submitted: Cases, deaths
Model name: Karlen
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Hospitalization assumptions: The model uses state-specific hospitalization data. New hospitalizations are estimated from these data, or from the estimated number of new infections that will occur in each location.
Methods: Discrete time difference equations
Forecasts submitted: Cases, hospitalizations, deaths
Model name: LNQ
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Machine learning
Forecasts submitted: Cases, deaths
Model name: Prolix
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Offsets obtained by correlations, best linear approximation of reproduction rates (using vaccination approximation) by least euclidean distance, and linear prediction.
Forecasts submitted: Cases, hospitalizations, and deaths
Model name: LSHTM
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: This forecast is an ensemble of three different models: A time-varying reproductive number-base model, a time series model based on numbers of deaths, and a time series model based on numbers of cases and deaths.
Forecasts submitted: Deaths
Model name: LANL
Intervention assumptions: This model assumes interventions in place on the first day of the forecast will remain in place for the next four weeks.
Hospitalization Assumptions: State demographics and age-group symptomatic case hospitalization rates are used to estimate the daily number of hospitalizations, based on estimates of the total number of infections.
Methods: Statistical dynamical growth model accounting for population susceptibility
Forecasts submitted: Cases, hospitalizations, deaths
Model name: LANL_NAU
Intervention assumptions: This model assumes interventions in place on the first day of the forecast will remain in place for the next four weeks.
Hospitalization Assumptions: None
Methods: Compartmental model consisting of ordinary differential equations (ODEs) describing the dynamics of 40 populations (state variables). The model quantified the impact of non-pharmaceutical interventions on COVID-19 transmission dynamics, and captures vaccination of susceptible and recovered persons and infected non-quarantined persons without symptoms at a time-varying per capita rate.
Forecasts submitted: Cases
Model name: LUcompUncertLab
Intervention assumptions: None
Hospitalization Assumptions: None
Methods: A Bayesian Vector Auto Regression model
Forecasts submitted: Hospitalizations, deaths
Model name: Masaryk
Intervention assumptions: The projections assume that current interventions will remain in place indefinitely.
Methods: An auto-regressive model
Forecasts submitted: Cases, deaths
Model name: MIT-Cassandra
Intervention Assumptions: The projections assume that current interventions will remain in place indefinitely.
Methods: This forecast is an ensemble of four different models: a Markov Decision Processes model, two different time series models, and C-SEIRD model
Forecasts submitted: Cases
Model name: MIT-CovAlliance
Intervention Assumptions: The projections assume that current interventions will remain in place indefinitely.
Methods: SIR model
Forecasts submitted: Cases, deaths
Model name: MIT-ISOLAT
Intervention Assumptions: The projections assume that current interventions will remain in place indefinitely.
Methods: Mixture model
Forecasts submitted: Cases, deaths
Model name: MIT-LCP
Intervention Assumptions: The projections assume that current interventions will remain in place indefinitely.
Methods: Machine learning
Forecasts submitted: Deaths
Model name: MIT-ORC
Intervention Assumptions: The projections assume that interventions will be reinstated if transmission reaches certain thresholds.
Methods: SEIR model
Forecasts submitted: Cases, deaths
Model name: Microsoft (formerly: MSRA)
Intervention Assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: SEIR model on a spatiotemporal network
Forecasts submitted: Cases, deaths
Model name: MOBS
Intervention assumptions: The projections assume that social distancing policies in place at the date of calibration are extended for the future weeks.
Methods: Metapopulation, age-structured SLIR model
Forecasts submitted: Hospitalizations, deaths
Model names:
Intervention assumptions: These forecasts assume that population-level mobility is a reliable proxy for adherence to social distancing, and that recent trends in mobility will continue over the coming weeks.
Methods:
- NotreDame-Mobility: SEIR model fit to data on deaths, test positivity, and population mobility
- NotreDame-Fred: Agent-based model
Forecasts submitted: Deaths
Model name: Oliver Wyman
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Time-dependent SIR model for detected and undetected cases
Forecasts submitted: Cases, deaths
Model name: OneQuietNight
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Machine learning models.
Forecasts submitted: Cases
Model name: PandemicCentral
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Random forest machine learning model
Forecasts submitted: Cases
Model name: PSI
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Stochastic SEIRX model
Forecasts submitted: Deaths
Model name: PSI-DICE
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: SEIR model
Forecasts submitted: Hospitalizations
Model name: QJHong
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Machine learning using case data and mobility data
Forecasts submitted: Cases, deaths
Model name: RPI-UW
Intervention assumptions: These projections calibrate the rate of transmission to the average rate of population mobility since the start of the epidemic and assume that this relationship will not change in the next four weeks.
Methods: SIR model fit to mobility data
Forecasts submitted: Deaths
Model name: ESG
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Fitting reported data to multiple skewed gaussian distributions
Forecasts submitted: Cases, deaths
Model name: SignatureScience
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Autoregressive time-series model
Forecasts submitted: Cases, deaths
Model name: UpstateSU
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Machine learning
Forecasts submitted: Cases, deaths
Model name: STH
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Statistical growth model
Forecasts submitted: Deaths
Model name: CovidComplete
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: Statistical prediction model
Forecasts submitted: Deaths
Model name: TTU
Intervention assumptions: This model assumes that the effects of interventions are reflected in the observed data and will continue going forward.
Methods: SIR model
Forecasts submitted: Cases
Model name: UA
Intervention assumptions: This model assumes that current interventions will remain in effect for at least four weeks after the forecasts are made.
Methods: SIR model with data assimilation
Forecasts submitted: Deaths
Model name: UCLA
Intervention assumptions: This model assumes that contact rates will increase as states reopen. The increase in contact rates is calculated for each state.
Hospitalization Assumptions: The number of new hospitalizations per day are estimated from the number of infections, using state-specific hospitalization rates.
Methods: Modified SEIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: UCM
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: SEIR model
Forecasts submitted: Deaths
Model name: UCSD-NEU
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Age-structured metapopulation model with deep learning
Forecasts submitted: Deaths
Model name: UCSB
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Hospitalization Assumptions: The number of new hospitalizations per day are estimated from the number of infections, using state-specific hospitalization rates.
Methods: An attention mechanism (deep learning) time series model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: UCF
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: SEIR model informed with with ensemble neural networks
Forecasts submitted: Cases
University of Chicago (forecasts for Illinois only)
Model name: UChicago
Intervention assumptions: These forecasts assume that transmission rate will increase by 10% when stay-at-home policies are lifted.
Methods: SEIR model
Forecasts submitted: Deaths
Model name: CUBoulder
Intervention assumptions: The projections assume that social distancing policies in place at the date of calibration are extended for the future weeks.
Methods: Our model is a stacked LSTM deep learning-based model using a multivariate time series input with a temporal lag of 9 weeks.
Forecasts submitted: Cases
Model name: UGA-CEID
Intervention assumptions: The projections assume that social distancing policies in place at the date of calibration are extended for the future weeks.
Methods: Statistical random walk model
Forecasts submitted: Cases and deaths
University of Geneva / Swiss Data Science Center (one-week ahead forecasts only)
Model name: Geneva
Intervention assumptions: The projections assume that social distancing policies in place at the date of calibration are extended for the future weeks.
Methods: Exponential and linear statistical models fit to the recent growth rate of cumulative deaths
Forecasts submitted: Cases, deaths
Model names: UMass-GBQ, UMass-MB, Ensemble, UMass-TE, UMass-sarix
Intervention assumptions:
- UMass-GBQ: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
- UMass-MB: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
- Ensemble: The ensemble forecasts include all submitted forecasts, derived from models that assume certain social distancing measures will continue and models that assume those measures will not continue.
- UMass-TE: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
- UMass-sarix: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
Methods:
- UMass-GBQ: Gradient boosting for the median, followed by conformal prediction to get quantiles.
- UMass-MB: Mechanistic Bayesian compartment model.
- Ensemble: The ensemble is a combination of 4 to 20 models, depending on the availability of forecasts for each location. To ensure consistency, the ensemble includes only models with 4 week-ahead forecasts.
- UMass-TE: Trends Ensemble model.
- UMass-sarix: Equally weighted ensemble of simple time-series baseline models. As of 2022-04-25, uses only past hospitalizations, no covariates.
Forecasts submitted: Cases, hospitalizations, deaths
Model name: UM
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Ridge regression
Forecasts submitted: Cases, deaths
Model name: USC
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Hospitalization assumptions: The number of new hospitalizations per day are estimated from the number of infections.
Methods: SIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: UT
Intervention assumptions: This model estimates the extent of social distancing using geolocation data from mobile phones and assumes that the extent of social distancing does not change during the period of forecasting. The model is designed to predict confirmed COVID-19 deaths resulting from only a single wave of transmission.
Methods: Nonlinear Bayesian hierarchical regression with a negative-binomial model for daily variation in death rates
Forecasts submitted: Deaths
Model name: UVA
Intervention assumptions: There are two different assumptions influencing this ensemble model. Two of the three models assume that the effects of interventions are reflected in the observed data and will continue going forward, while the third model assumes that interventions will change in the future.
Methods: This forecast is an ensemble of three different models: An auto-regressive model, a machine learning (long short-memory) model, and a SEIR model.
Forecasts submitted: Cases
Model name: ERDC
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Hospitalization assumptions: The number of new hospitalizations per day are estimated from the number of infections, using state-specific hospitalization rates.
Methods: SEIR model
Forecasts submitted: Cases, hospitalizations, deaths
Model name: Walmart
Intervention assumptions: These projections do not make any specific assumptions about which interventions have been implemented or will remain in place.
Methods: Logistic growth
Forecasts submitted: Deaths
Model name: Wadhwani
Intervention assumptions: These projections assume that current interventions will not change during the forecasted period.
Methods: Bayesian SEIR model
Forecasts submitted: Cases, deaths
Model name: YYG
Intervention assumptions: The model accounts for individual state-by-state re-openings and responses to increasing cases and their impact on infections and deaths.
Methods: SEIR model with machine learning for parameter estimation
Forecasts submitted: Deaths