Member Junction
    Preparing search index...

    Prediction Unit and pipeline fixtures for the gates. Later provisioners are added here too.

    Index

    Constructors

    Methods

    • The catalog algorithm a scenario pipeline of a task trains: logistic regression for a class, ridge for a continuous value, and a random forest (Poisson splits) for a count or non-negative value, which the trainer fits with neither GLM (its Poisson and Tweedie GLMs are baselines only).

      Parameters

      Returns string

    • The scenario table's feature columns: every field except keys, the target, the planted-truth columns and the date column T (always: a scenario without dated rows still has a T column, NULL on every row) and dateColumn.

      Parameters

      Returns readonly string[]

    • A cross-sectional planted scenario (no dates): one decision moment for every row; groupKey groups its rows (folds, new-entity report) when the scenario has several rows per entity.

      Parameters

      Returns UnitDraft

    • The pstruth entity registered for a base table (names come from the test schema's CodeGen config, so they are looked up).

      Parameters

      Returns string

    • The per-world table prefix: 'W5-censor' → 'W5' (tables W5_Person, W5_MembershipYear).

      Parameters

      • world: string

      Returns string