quantlab.model.predefined.double_ensemble

Qlib’s DoubleEnsemble benchmark model.

DoubleEnsembleRegressor reproduces Qlib’s DEnsembleModel (qlib/contrib/model/double_ensemble.py; Zhang et al., “DoubleEnsemble: A New Ensemble Method Based on Sample Reweighting and Feature Selection for Financial Data Analysis”, ICDM 2020) with LightGBM sub-models and its Alpha158 benchmark parameters (workflow_config_doubleensemble_Alpha158.yaml). LightGBM ships in the gbdt extra.

Classes

DoubleEnsembleBoosters	The fitted sub-models, the feature columns of each and their weights.
DoubleEnsembleRegressor	Qlib's DoubleEnsemble of LightGBM sub-models on the rank target's rows.

Functions

pct_rank	pandas' rank(pct=True): the average rank over the count, ties averaged.
sample_weights	Qlib's sample re-weighting (DEnsembleModel.sample_reweight).
select_features	Qlib's feature selection draw (DEnsembleModel.feature_selection) from the g-values.