libuplift.metrics.bins#

Measures based on comparing treatment and control statistics within bins, such as quantiles.

Functions#

`iter_quantiles`(scores, trt, n_trt[, n, joint, ...])	Iterate simultaneously over quantiles of score vectors for all
`QMSE`(y_true, y_pred, trt[, n_trt, sample_weight, ...])	The per-quantile MSE measure by Rudaś, Jaroszewicz.
`QMSE_j`(y_true, y_pred, trt[, n_trt, sample_weight, ...])
`EUCE`(y_true, y_pred, trt[, n_trt, sample_weight, ...])	The EUCE measure by Nyberg and Klami.
`MUCE`(y_true, y_pred, trt[, n_trt, sample_weight, ...])	The MUCE measure by Nyberg and Klami.

libuplift.metrics.bins.iter_quantiles(scores, trt, n_trt, n=10, joint=False, sample_weight=None)[source]#

Iterate simultaneously over quantiles of score vectors for all treatments.

Returns a generator which, for each quantile, returns a list of index arrays for scores within each treatment.

If joint is True, quantiles are computed jointly for all treatments.

If sample_weight is not None, weighted quantiles are used.

libuplift.metrics.bins.QMSE(y_true, y_pred, trt, n_trt=None, sample_weight=None, n_bins=10, allow_nans=False, joint_quantiles=False)[source]#: The per-quantile MSE measure by Rudaś, Jaroszewicz.

libuplift.metrics.bins.QMSE_j(y_true, y_pred, trt, n_trt=None, sample_weight=None, n_bins=10, allow_nans=False)[source]#

libuplift.metrics.bins.EUCE(y_true, y_pred, trt, n_trt=None, sample_weight=None, n_bins=100, allow_nans=False, joint_quantiles=True)[source]#: The EUCE measure by Nyberg and Klami.

libuplift.metrics.bins.MUCE(y_true, y_pred, trt, n_trt=None, sample_weight=None, n_bins=100, allow_nans=False, joint_quantiles=True)[source]#: The MUCE measure by Nyberg and Klami.