On this page we explain the fundamental ideas behind Zoltar's envisioning of representing forecasts in a challenge.
- Projects: As mentioned in Manging Projects, projects are the central organizing concept in Zoltar. They define the units, targets and time zeros that forecasts are relative to, and contain the forecast models that teams use to represent a forecasting challenge.
- Forecast models: A model is the representation of code that generates forecasts. It has zero or one forecasts associated with the project's time zeros.
- Forecasts: A forecast consists of a set of predictions of the various types supported by Zoltar. Forecast versioning is supported via each forecast having an
issued_atfield, which defaults to when the forecast was uploaded to Zoltar. (Users with advanced permission can change that field.) See the Data Model and Forecast Versions pages for details.
- Units: Units represent distinct entities for which predictions are made, for example, if forecasts are made for multiple different locations, the locations are the units.
- Targets: Targets are the fundamental data structure of a forecast, and as such have a separate page documenting them.
- Time zeros: Because the forecasting field does not have standard terminology, we have settled on the following two concepts for this application. Note that some time zeros are tagged as starting a season, specifying the season's name, which helps to segment the time zeros.
- Time zero: The date from which a forecast originates and to which targets are relative (i.e. a "2-week-ahead forecast" is two weeks ahead of the time-zero). Every forecast is for a time zero.
- Data version date: (Optional) The latest date at which any data source used for the forecasts should be considered. Can be used externally to recreate model results by "rolling back" the core data to a particular state.
- Truth: Zoltar supports uploading ground truth data for projects. Please see this page for details.