def test_does_quack_expected_values_mixin_negative(self):
NoExpectedValues = type('InvalidExpectedValues', (BaseValid, FeatureImportanceValid, GlobalValid), {})
assert not ExpectedValuesMixin._does_quack(NoExpectedValues())
class ExpectedValuesNone(object):
@property
def expected_values(self):
return None
ExpectedValuesNoneMixin = type('InvalidExpectedValues', (ExpectedValuesNone,), {})
assert not ExpectedValuesMixin._does_quack(ExpectedValuesNoneMixin())
def create_global(self,
local_importance_values,
evaluation_examples=None,
expected_values=None,
include_local=True,
batch_size=Defaults.DEFAULT_BATCH_SIZE):
"""Create a global explanation from the list of local feature importance values.
:param local_importance_values: The feature importance values.
:type local_importance_values: numpy.array or scipy.sparse.csr_matrix or list[scipy.sparse.csr_matrix]
:param evaluation_examples: A matrix of feature vector examples (# examples x # features) on which
to explain the model's output.
:type evaluation_examples: numpy.array or pandas.DataFrame or scipy.sparse.csr_matrix
:param expected_values: The expected values of the model.
:type expected_values: numpy.array
:param include_local: Include the local explanations in the returned global explanation.
If include_local is False, will stream the local explanations to aggregate to global.
:type include_local: bool
:param batch_size: If include_local is False, specifies the batch size for aggregating
local explanations to global.
:type batch_size: int
"""
local_explanation = self.create_local(local_importance_values,
evaluation_examples,
expected_values)
kwargs = {ExplainParams.METHOD: self.method}
kwargs[ExplainParams.FEATURES] = self.features
if ExpectedValuesMixin._does_quack(local_explanation):
kwargs[ExplainParams.
EXPECTED_VALUES] = local_explanation.expected_values
if include_local:
kwargs[ExplainParams.LOCAL_EXPLANATION] = local_explanation
# Aggregate local explanation to global
return _aggregate_global_from_local_explanation(**kwargs)
else:
if ExplainParams.CLASSIFICATION in kwargs:
if kwargs[ExplainParams.CLASSIFICATION]:
model_task = ModelTask.Classification
else:
model_task = ModelTask.Regression
else:
model_task = ModelTask.Unknown
kwargs = _aggregate_streamed_local_explanations(
self, evaluation_examples, model_task, self.features,
batch_size, **kwargs)
return _create_global_explanation(**kwargs)
def test_does_quack_expected_values_mixin(self):
ValidExpectedValues = type('ValidExpectedValues', (ExpectedValuesValid,), {})
assert ExpectedValuesMixin._does_quack(ValidExpectedValues())