tic.interpret package

Submodules

tic.interpret.feature_importance module

tic.interpret.feature_importance.explain_global(clf: sklearn.base.BaseEstimator, feature_names: List, absolute_values: Optional[bool] = False, sort: Optional[bool] = True, num_features: Optional[int] = None)

Uses built-in methods from the classifier object to extract feature importances in form of Gini coefficients (for tree based models) and coefficients (for linear models).

Args:

clf: fitted classifier object as input feature_names: feature names as fed to the fitting function of clf absolute_values: whether absolute values should be returned (exp) sort: whether the importances should be sorted num_features: how many importances should be reported

Returns:

list of feature importances with the given configuration

tic.interpret.lime module

tic.interpret.lime.explain_global(clf: sklearn.base.BaseEstimator, training_data: List, feature_names: List, class_names: List, explainer_kwargs: Optional[Dict] = {}, explanation_kwargs: Optional[Dict] = {})

Creates an explainer and creates a submodular pick. Returns the explanation with the highest coverage.

Args:

clf : Fitted classifier from sklearn training_data: data that was used to train the classifier feature_names: name of features of dataset class_names: names of class labels explainer_kwargs: Keyword args passed during explainer initialization explanation_kwargs: Keyword args passed for submodular pick

Returns:

Enriched explanation with highest coverage including figure

tic.interpret.lime.explain_local(clf: sklearn.base.BaseEstimator, instance: List, training_data: List, feature_names: List, class_names: List, explainer_kwargs: Optional[Dict] = {}, explanation_kwargs: Optional[Dict] = {})

Creates an explainer and explains the given instance using LIME.

Args:

clf : Fitted classifier from sklearn instance: instance to explain training_data: data that was used to train the classifier feature_names: name of features of dataset class_names: names of class labels explainer_kwargs: Keyword args passed during explainer initialization explanation_kwargs: Keyword args passed for explanation

Returns:

Enriched explanation including figure

tic.interpret.shap module

tic.interpret.shap.explain_global(clf: sklearn.base.BaseEstimator, X_train: pandas.core.frame.DataFrame, X_test: pandas.core.frame.DataFrame, class_names: List, sample_size: Optional[int] = 100, explainer_kwargs: Optional[Dict] = {}, explanation_kwargs: Optional[Dict] = {})

Creates an explainer and explanations for a given dataset using SHAP.

Args:

clf : Fitted classifier from sklearn X_train: data that was used to train the classifier X_test: data that should be explained class_names: names of class labels sample_size: how many data points are used to create the SHAP values explainer_kwargs: Keyword args passed during explainer initialization explanation_kwargs: Keyword args passed for explanation

Returns:

Enriched SHAP explanation including interactive figure

tic.interpret.shap.explain_local(clf: sklearn.base.BaseEstimator, X_train: pandas.core.frame.DataFrame, instance: pandas.core.series.Series, class_names: List, sample_size: Optional[int] = 100, explainer_kwargs: Optional[Dict] = {}, explanation_kwargs: Optional[Dict] = {})

Creates an explainer and explains the given instance using SHAP.

Args:

clf : Fitted classifier from sklearn X_train: data that was used to train the classifier instance: instance to explain class_names: names of class labels sample_size: how many data points are used to create the SHAP values explainer_kwargs: Keyword args passed during explainer initialization explanation_kwargs: Keyword args passed for explanation

Returns:

Enriched SHAP explanation including figure

tic.interpret.surrogate module

tic.interpret.surrogate.explain_global(clf: sklearn.base.BaseEstimator, surrogate_type: str, X_train: pandas.core.frame.DataFrame, num_features: int, surrogate_kwargs: Optional[Dict] = {}, feature_importances_kwargs: Optional[Dict] = {})

Creates a surrogate model that mimics the behavior of the original model.

Args:

clf : Fitted classifier from sklearn surrogate_type: Type of surrogate model, linear or tree X_train: DataFrame with the original training data num_features: how many feature importances should be returned surrogate_kwargs: Keyword args passed during surrogate initialization feature_importances_kwargs: Kwargs passed for feature importance method

Returns:

Surrogate model with its feature importances

Module contents