Example forecast hub data — forecast

The hubExamples package provides three data sets that contain example model output and target data for an example forecast hub: forecast_outputs, forecast_target_ts, and forecast_target_observations.

forecast_outputs contains example forecast data that represents model outputs from a forecast hub with predictions for three influenza-related targets (wk inc flu hosp, wk flu hops rate category, and wk flu hosp rate) for two reference dates in 2022. This dataset reflect forecast outputs as they look when retrieved from a hub via the hubData package (which is slightly different than they look when originally submitted by modelers).

forecast_target_ts contains time series target data from a hub that predicts influenza-related targets.

forecast_target_observations contains target data that represents the source of "truth" that model output data will be scored against. This example represents influenza-related targets.

Usage

forecast_outputs

forecast_target_ts

forecast_target_observations

Format

`forecast_outputs`

A data frame with 5,424 rows and 9 columns:

location: FIPS code identifying a location
reference_date: the starting point of the forecast in yyyy-mm-dd format
horizon: number of time units ahead being forecasted relative to the reference_date, in units of weeks
target_end_date: the date of occurrence of the outcome of interest in yyyy-mm-dd format; this can be calculated directly from the reference_date and horizon as follows: target_end_date = reference_date + 7*horizon
target: a unique identifier for the target
output_type: the type of representation of the prediction
output_type_id: more identifying information specific to the output type; output_type_id is not relevant for every kind of output_type (for example, hubs will not expect output_type_id values when the output_type is mean or median
value: the model’s prediction
model_id: the name of the model

`forecast_target_ts`

A data frame with 10,255 rows and 3 columns:

date: the date of the target observation in yyyy-mm-dd format
location: FIPS code identifying a location
observation: a count of hospital admissions in the given location in the week ending on the given date.

`forecast_target_observations`

A data frame with 198,485 rows and 6 columns:

location: FIPS code identifying a location
target_end_date: the target's observation date in yyyy-mm-dd format; this is used to match on the target_end_date field in model output data submitted to the hub
target: a unique identifier for the target
output_type: the type of representation of the prediction
output_type_id: more identifying information specific to the output type; as in the model output data, output_type_id is not relevant for output_type of mean and median; target data that represents quantile output_type will not have an output_type_id.
observation: the observed value of the target

Source

https://github.com/hubverse-org/example-complex-forecast-hub/