rusboost

era_pl.rusboost(
    data,
    features,
    target,
    *,
    size=30,
    rus=True,
    learn_rate=1.0,
    maxdepth=None,
    minbucket=None,
    ir=1.0,
    random_state=2021,
)

Fit a binary boosting model using at most size learner attempts.

A learner that correctly classifies the full training data replaces all earlier learners and receives voting weight 1. Unusable learners stop AdaBoost (rus=False); RUSBoost skips them and tries another sample. Raise ValueError if no usable learner is fitted. Single-class data retain the ConstantProbModel fallback.