explainy.core package#

Submodules#

explainy.core.explanation module#

class explainy.core.explanation.Explanation(score_text: str, method_text: str, natural_language_text: str, separator: str = '\n')[source]#

Bases: object

Explanation class

get_explanation(separator: str = '\n') → str | Tuple[str, str, str][source]#

Get the final explanation as a string or tuple

Args:

separator (str, optional): Separator to join the sub-explanations. Defaults to “

“.

Returns:

Union[str, Tuple[str, str, str]]: final explanation

print_output(separator: str = '\n') → None[source]#

explainy.core.explanation_base module#

class explainy.core.explanation_base.ExplanationBase(model: ClassifierMixin | RegressorMixin, config: Dict[str, str] | None = None)[source]#

Bases: ABC

define_explanation_placeholder(natural_language_text_empty: str, method_text_empty: str, sentence_text_empty: str) → None[source]#

Set the explanation text, if defined else load it from defaults

Parameters:

  • natural_language_text_empty (str) – natural language explanation placeholder

  • method_text_empty (str) – method placeholder

  • sentence_text_empty (str) – sentence text placeholder

get_description_text() → str[source]#

Generate the description of the explanation method.

Example

This is a SHAP explanation, it creates local and non-contrastive explanations.

Returns:

return the explanation method description

Return type:

str

get_feature_names(X: DataFrame | ndarray) → List[str][source]#

Get the feature names based on the given dataset

Parameters:

X (Union[pd.DataFrame, np.array]) – features dataset

Returns:

list of feature names

Return type:

List[str]

abstract get_feature_values()[source]#

Get the feature values

get_method_text() → str[source]#

Generate the output of the method explanation.

get_model_text() → str[source]#

Generate text the explains the used machine learning model (wip)

Returns:

return the description of the machine learning model

Return type:

str

get_natural_language_text() → str[source]#

Generate the output of the explanation in natural language.

Returns:

return the natural_language_text explanation

Return type:

str

get_number_of_features(number_of_features: int) → int[source]#

Set the number of features based on the defined number and the max number of features

Parameters:

number_of_features (int) – number_of_features as input

Returns:

number_of_features considering the max number of dataset features

Return type:

int

get_plot_name(sample_name: str | None = None) → str[source]#

Get the name of the plot

Parameters:

sample_name (str, optional) – name of the sample. Defaults to None.

Returns:

return the name of the plot

Return type:

str

get_prediction(sample_index: int) → float[source]#

Get the model prediction

Parameters:

sample_index (int) – sample_index for a which a predction shall be made

Returns:

predction of the model for that sample

Return type:

float

get_sample_name(sample_index: int, sample_name: str | None = None) → str[source]#

Determine the name of the sample, if no sample_name provide, use the sample_index

Parameters:

  • sample_index (int) – index of the sample

  • sample_name (str, optional) – name of the sample. Defaults to None.

Returns:

name of the sample

Return type:

str

get_score_text() → str[source]#

Generate the text explaining the prediction score of the sample

Returns:

return the score_text for the sample.

Return type:

str

get_sentences() → str[source]#

Generate the output sentences of the explanation.

importance() → DataFrame[source]#

Get the feature importance

abstract plot(sample_index: int, kind: str) → None[source]#

Plot the feature importance

save(sample_index: int, sample_name: str | None = None) → None[source]#

Save the explanations to a csv file, save the plots

Parameters:

  • sample_index (int) – [description]

  • sample_name (str, optional) – name of the sample. Defaults to None.

save_csv(sample_index: int) → None[source]#

Save the explanation to a csv. The columns contain the method_text, the natural_language_text, the name of the plot and the predicted value. The index is the Entry ID.

Parameters:

sample_index (int) – index of the sample

Returns:

None.

set_paths() → None[source]#

Set the paths where the output should be saved

Module contents#