GBLinearForecaster#

class openstef_models.models.forecasting.gblinear_forecaster.GBLinearForecaster(**data: Any) → None[source]

Bases: Forecaster, ExplainableForecaster, ContributionsMixin

GBLinear-based forecaster for probabilistic energy forecasting.

Implements gradient boosted linear models using XGBoost’s gblinear booster for multi-quantile forecasting. Unlike tree-based models, this linear approach does not suffer from extrapolation issues and provides better performance for time series data where predictions outside the training range are required.

The forecaster uses linear models with gradient boosting optimization, making it particularly suitable for forecasting scenarios where the underlying relationships are approximately linear or when avoiding extrapolation artifacts is critical. This approach provides well-calibrated uncertainty estimates while maintaining computational efficiency and interpretability.

Invariants

fit() must be called before predict() to train the model
Configuration quantiles determine the number of prediction outputs
Model state is preserved across predict() calls after fitting
Input features must match training data structure during prediction

Example

Basic forecasting workflow

>>> from datetime import timedelta
>>> from openstef_core.types import LeadTime, Quantile
>>> forecaster = GBLinearForecaster(
...     quantiles=[Quantile(0.1), Quantile(0.5), Quantile(0.9)],
...     horizons=[LeadTime(timedelta(hours=1))],
...     hyperparams=GBLinearHyperParams(
...         learning_rate=0.1,
...         reg_alpha=0.1,
...         reg_lambda=1.0,
...     ),
... )
>>> forecaster.fit(training_data)
>>> predictions = forecaster.predict(test_data)

Note

XGBoost dependency is optional and must be installed separately. The model automatically handles multi-quantile output using quantile regression and is optimized for energy forecasting applications where linear relationships dominate and extrapolation beyond training data is required.