sapient bodhiForecast
Help

Run creation help

Choose a guided form or API-compatible JSON configuration, then review the schema requirements and pipeline concepts used when starting forecast runs.

Browse help topics and this page

Help topics

Configure a run

  • Choose form or JSON

    Use the guided form to configure a run step by step. JSON configuration can load a starter template from the current Forecast API contract. Keys rejected by the API must be removed before submission; keys the API would silently ignore are identified but do not block submission. If the contract is unavailable, the existing starter JSON and backend validation still apply.

  • Enter column lists

    Add each column as a separate item by pressing Enter or typing a comma. You can also paste comma-separated or line-separated values and remove individual items.

  • Reuse configuration

    Reuse configuration starts a new run with settings from an earlier run. Fields the current API no longer supports are removed from the copied configuration and listed for review; the saved run is not changed. Review source credentials, dates, and outputs before submission.

  • Create an experiment during submission

    A newly created experiment is selected automatically. A source or parent run's experiment is suggested when the relationship is unambiguous, but you can change it.

Schema terminology

TermRequirementExamplesMeaning
Target Column(s)
Required
sales
temperature, humidity
Column name(s) containing observations of the endogenous variable(s) for which forecasts are required.
Timestamp Column
Required
pos_date
utc_time
Column name containing temporal information for each row in the dataset.
Identifier Column(s)
Optional
product_id
state, city
Column name(s) containing hierarchical information used to identify a single time series.
Static Column(s)
Optional
product_category, product_height, product_weight, product_depth
country
Column name(s) containing exogenous feature values that remain fixed for a single time series.
Known Column(s)
Optional
discount_rate
season
Column name(s) containing exogenous feature values that vary over time and are known or controllable during the forecast horizon.
Unknown Column(s)
Optional
price, GDP
wind_speed
Column name(s) containing exogenous feature values that vary over time when no prior information is available for the forecast horizon.

Data requirements

  1. Data must be in long format. Each row represents one observation for all endogenous and exogenous features for one time series at one point in time.
  2. Target variables must be real-valued.
  3. The timestamp column can contain a date or a date and time. All observations must use the same format, preferably ISO 8601.
  4. Identifier columns are optional when the dataset contains only one time series.
  5. Static, known, and unknown variables can be categorical or real-valued.

Orchestration terminology

TermMeaningDetails
PipelineA sequence of Bodhi Forecast SDK capabilities invoked for a specific purpose.
  1. Training: Uses past target and feature data to create a trained model.
  2. Prediction: Uses a trained model and configured future feature data to create target predictions.
  3. Complete: Trains a model and creates predictions in one run using the configured historical and future data.
  4. Past Explainability: Uses past target and feature data to generate feature importances.
  5. Future Evaluation: Evaluates forecasts against actual future targets and can provide optional importance results.
RunA single execution of one Bodhi Forecast pipeline.A run can be started directly or created as part of a project.
ExperimentA collection of runs.Typically groups runs that use related datasets or configurations.