Model: Protea - Dynamic Harmonic Model with ARIMA errors API

    Name: Dynamic Harmonic Model with ARIMA errors
    Abbreviation: Springbok
    Owner: kgle
    Project: CDC Real-time Forecast
    Team name: Protea
    Description: Team name: Protea
    Team members: Craig J. McGowan (contact), Alysse J. Kowalski
    Data source(s): ilinet, google trends, who-nrevss lab data
    Methods: A dynamic harmonic regression model is fit separately for each region in
    sequential steps, with Fourier terms to capture seasonality and multiple
    possible covariates. At each sequential step, cross-validated models are fit
    using each of the seasons from 2010/2011 through 2018/2019 and scored using
    CDC single bin scoring rules. First, the number of Fourier terms (ranging
    from 1 to 12) is selected. The non-seasonal ARIMA structure of the error terms
    is selected second, and potential covariates are tested for inclusion last.
    Covariates include influenza virus subtype 6 week rolling average percentage,
    national Google Trends data, and regional Google Trends data (using the most
    populous state in a given HHS region) along with combinations of those
    covariates. Final CV model scores were examined and the top performing model
    structure was chosen for each location. Forecasts are simulated from iterating
    one-step-ahead preditions for the remainder of the season, and the occurence
    of seasonal targets is calculated from these predictions. Observed values for
    the current season in the simulation are adjusted for backfill by sampling
    from a distribution of prior observed outcomes for that lag/week combination.
    Predicted probabilities are calculated from the observed values across
    multiple simulations. For making prospective predictions, models were fit
    using only data from prior seasons, and forecasts were made using data that
    would have been available at the time of the forecast. In the case of
    covariates, predicted values of the covariates based on prior data were used
    to generate forecasts. For Google Trends data, no information on backfill is
    available and the data are assumed to be free of backfill. For influenza
    subtypes, no information on backfill is available, so we use cumulative
    percentages of each influenza subtype up to a given week in the season, which
    is less susceptible to backfill effects than weekly measures. All code used
    in estimation and prediction is available at
    https://github.com/craigjmcgowan/FluForecast
    Home: https://github.com/FluSightNetwork/cdc-flusight-ensemble/tree/master/model-forecasts/real-time-component-models/Protea_Springbok
    Auxiliary data: (No auxiliary data)

    Forecasts (33)

    Timezero Data Source Upload Date
    2017-10-23 (No data)
    2017-10-30 (No data)
    2017-11-06 (No data)
    2017-11-13 (No data)
    2017-11-20 (No data)
    2017-11-27 (No data)
    2017-12-04 (No data)
    2017-12-11 (No data)
    2017-12-18 (No data)
    2017-12-25 (No data)
    2018-01-01 (No data)
    2018-01-08 (No data)
    2018-01-15 (No data)
    2018-01-22 (No data)
    2018-01-29 (No data)
    2018-02-05 (No data)
    2018-02-12 (No data)
    2018-02-19 (No data)
    2018-02-26 (No data)
    2018-03-05 (No data)
    2018-03-12 (No data)
    2018-03-19 (No data)
    2018-03-26 (No data)
    2018-04-02 (No data)
    2018-04-09 (No data)
    2018-04-16 (No data)
    2018-04-23 (No data)
    2018-04-30 (No data)
    2018-10-08 (No data)
    2018-10-15 EW42-2018-Protea_Springbok.csv 2019-10-22 07:32:53
    2018-10-22 EW43-2018-Protea_Springbok.csv 2019-10-22 07:33:25
    2018-10-29 EW44-2018-Protea_Springbok.csv 2019-10-22 07:34:03
    2018-11-05 EW45-2018-Protea_Springbok.csv 2019-10-22 07:34:40
    2018-11-12 EW46-2018-Protea_Springbok.csv 2019-10-22 07:35:23
    2018-11-19 EW47-2018-Protea_Springbok.csv 2019-10-22 07:35:59
    2018-11-26 EW48-2018-Protea_Springbok.csv 2019-10-22 07:36:41
    2018-12-03 EW49-2018-Protea_Springbok.csv 2019-10-22 07:37:19
    2018-12-10 EW50-2018-Protea_Springbok.csv 2019-10-22 07:38:05
    2018-12-17 EW51-2018-Protea_Springbok.csv 2019-10-22 07:38:49
    2018-12-24 EW52-2018-Protea_Springbok.csv 2019-10-22 07:39:27
    2018-12-31 EW01-2019-Protea_Springbok.csv 2019-10-22 07:40:01
    2019-01-07 EW02-2019-Protea_Springbok.csv 2019-10-22 07:40:41
    2019-01-14 EW03-2019-Protea_Springbok.csv 2019-10-22 07:41:24
    2019-01-21 EW04-2019-Protea_Springbok.csv 2019-10-22 07:42:03
    2019-01-28 EW05-2019-Protea_Springbok.csv 2019-10-22 07:42:39
    2019-02-04 EW06-2019-Protea_Springbok.csv 2019-10-22 07:43:33
    2019-02-11 EW07-2019-Protea_Springbok.csv 2019-10-22 07:44:19
    2019-02-18 EW08-2019-Protea_Springbok.csv 2019-10-22 07:44:57
    2019-02-25 EW09-2019-Protea_Springbok.csv 2019-10-22 07:45:34
    2019-03-04 EW10-2019-Protea_Springbok.csv 2019-10-22 07:46:11
    2019-03-11 EW11-2019-Protea_Springbok.csv 2019-10-22 07:46:46
    2019-03-18 EW12-2019-Protea_Springbok.csv 2019-10-22 07:47:21
    2019-03-25 EW13-2019-Protea_Springbok.csv 2019-10-22 07:47:58
    2019-04-01 EW14-2019-Protea_Springbok.csv 2019-10-22 07:48:50
    2019-04-08 EW15-2019-Protea_Springbok.csv 2019-10-22 07:49:27
    2019-04-15 EW16-2019-Protea_Springbok.csv 2019-10-22 07:50:09
    2019-04-22 EW17-2019-Protea_Springbok.csv 2019-10-22 07:50:53
    2019-04-29 EW18-2019-Protea_Springbok.csv 2019-10-22 07:51:42
    2019-09-30 (No data)
    2019-10-07 EW41-2019-Protea_Springbok.csv 2019-10-31 03:18:59
    2019-10-14 EW42-2019-Protea_Springbok.csv 2019-10-31 03:19:33
    2019-10-21 EW43-2019-Protea_Springbok.csv 2019-11-06 11:45:45
    2019-10-28 EW44-2019-Protea_Springbok.csv 2019-11-15 08:47:46
    2019-11-04 (No data)
    2019-11-11 (No data)
    2019-11-18 (No data)
    2019-11-25 (No data)
    2019-12-02 (No data)
    2019-12-09 (No data)
    2019-12-16 (No data)
    2019-12-23 (No data)
    2019-12-30 (No data)
    2020-01-06 (No data)
    2020-01-13 (No data)
    2020-01-20 (No data)
    2020-01-27 (No data)
    2020-02-03 (No data)
    2020-02-10 (No data)
    2020-02-17 (No data)
    2020-02-24 (No data)
    2020-03-02 (No data)
    2020-03-09 (No data)
    2020-03-16 (No data)
    2020-03-23 (No data)
    2020-03-30 (No data)
    2020-04-06 (No data)
    2020-04-13 (No data)
    2020-04-20 (No data)
    2020-04-27 (No data)
    2020-05-04 (No data)
    2020-05-11 (No data)
    2020-05-18 (No data)