libuplift.meta.response#

‘Fake’ uplift models based on response classifiers.

Classes#

TreatmentUpliftClassifier

Predict uplift based on treatment classifiers.

ResponseUpliftClassifier

Predict uplift using a classifier built on full data.

ControlUpliftClassifier

Predict uplift based on a control classifier.

Module Contents#

class libuplift.meta.response.TreatmentUpliftClassifier(base_estimator=LogisticRegression(), reverse=False)[source]#

Bases: _ResponseModelBase

Predict uplift based on treatment classifiers.

Ignore control.

class libuplift.meta.response.ResponseUpliftClassifier(base_estimator=LogisticRegression(), reverse=False)[source]#

Bases: TreatmentUpliftClassifier

Predict uplift using a classifier built on full data.

Ignore causal nature of the data.

fit(X, y, trt, n_trt=None, sample_weight=None)[source]#
class libuplift.meta.response.ControlUpliftClassifier(base_estimator=LogisticRegression(), reverse=True)[source]#

Bases: TreatmentUpliftClassifier

Predict uplift based on a control classifier.

Ignore treatment data. If reverse is True lower classification scores are assumed to correspond to higher uplift.

fit(X, y, trt, n_trt=None, sample_weight=None)[source]#