facet.explanation.KernelExplainerFactory#

class facet.explanation.KernelExplainerFactory(*, link=None, l1_reg='num_features(10)', data_size_limit=100, **explainer_kwargs)[source]#

A factory constructing KernelExplainer instances.

Bases

FunctionExplainerFactory

Metaclasses

ABCMeta

Parameters

link (Optional[str]) – override the default link parameter (optional)
l1_reg (Optional[str]) – override the default l1_reg parameter of method shap_values(); pass None to use the default value used by shap_values() (optional)
data_size_limit (Optional[int]) – maximum number of observations to use as the background data set; larger data sets will be down-sampled using kmeans; don’t downsample if omitted (optional)
explainer_kwargs (Any) – additional keyword arguments to be passed to the explainer

Method summary

`make_explainer`	Construct a new `Explainer` to compute shap values.
`make_explainer_from_function`	Construct an explainer from a function.
`to_expression`	Render this object as an expression.

Attribute summary

`explains_raw_output`	`True` if explainers made by this factory explain raw model output, `False` otherwise.
`supports_shap_interaction_values`	`True` if explainers made by this factory allow for calculating SHAP interaction values, `False` otherwise.
`uses_background_dataset`	`True`, since function explainers typically use a background dataset
`explainer_kwargs`	Additional keyword arguments to be passed to the explainer constructor.

Definitions

make_explainer(model, data)#

Construct a new Explainer to compute shap values.

Parameters

model (Union[RegressorMixin, ClassifierMixin, Callable[[Union[Series, DataFrame, ndarray[Any, dtype[float64]]]], Union[Series, ndarray[Any, dtype[float64]], float]]]) – fitted learner for which to compute shap values
data (Optional[DataFrame]) – background dataset (optional)

Return type

BaseExplainer

Returns

the new explainer instance

make_explainer_from_function(model_fn, data)[source]#

Construct an explainer from a function.

Parameters

model_fn (Callable[[Union[Series, DataFrame, ndarray[Any, dtype[float64]]]], Union[Series, ndarray[Any, dtype[float64]], float]]) – the function representing the model
data (Optional[DataFrame]) – the background dataset

Return type

BaseExplainer

Returns

the explainer

to_expression()[source]#

Render this object as an expression.

property explains_raw_output: bool#

True if explainers made by this factory explain raw model output, False otherwise.

property supports_shap_interaction_values: bool#

True if explainers made by this factory allow for calculating SHAP interaction values, False otherwise.

property uses_background_dataset: bool#

True, since function explainers typically use a background dataset

facet.explanation.FunctionExplainerFactory

facet.explanation.PermutationExplainerFactory