SparkNpPermutationImportanceEstimator

class sparklightautoml.pipelines.selection.permutation_importance_based.SparkNpPermutationImportanceEstimator(random_state=42, computations_settings=None)[source]

Bases: SparkImportanceEstimator

Permutation importance based estimator.

Importance calculate, using random permutation of items in single column for each feature.

__init__(random_state=42, computations_settings=None)[source]

Parameters:: random_state (int) – seed for random generation of features permutation.

fit(train_valid=None, ml_algo=None, preds=None)[source]

Find importances for each feature in dataset.

Parameters:

train_valid (Optional[SparkBaseTrainValidIterator]) – Initial dataset iterator.
ml_algo (Optional[SparkTabularMLAlgo]) – Algorithm.
preds (Optional[SparkDataset]) – Predicted target values for validation dataset.