feature_explanations
=========================================================

.. py:module:: evalml.model_understanding.feature_explanations

.. autoapi-nested-parse::

   Human Readable Pipeline Explanations.



Module Contents
---------------



Functions
~~~~~~~~~

.. autoapisummary::
   :nosignatures:

   evalml.model_understanding.feature_explanations.get_influential_features
   evalml.model_understanding.feature_explanations.readable_explanation



Contents
~~~~~~~~~~~~~~~~~~~
.. py:function:: get_influential_features(imp_df, max_features=5, min_importance_threshold=0.05, linear_importance=False)

   Finds the most influential features as well as any detrimental features from a dataframe of feature importances.

   :param imp_df: DataFrame containing feature names and associated importances.
   :type imp_df: pd.DataFrame
   :param max_features: The maximum number of features to include in an explanation. Defaults to 5.
   :type max_features: int
   :param min_importance_threshold: The minimum percent of total importance a single feature can have to be considered important. Defaults to 0.05.
   :type min_importance_threshold: float
   :param linear_importance: When True, negative feature importances are not considered detrimental. Defaults to False.
   :type linear_importance: bool

   :returns: Lists of feature names corresponding to heavily influential, somewhat influential, and detrimental features, respectively.
   :rtype: (list, list, list)


.. py:function:: readable_explanation(pipeline, X=None, y=None, importance_method='permutation', max_features=5, min_importance_threshold=0.05, objective='auto')

   Outputs a human-readable explanation of trained pipeline behavior.

   :param pipeline: The pipeline to explain.
   :type pipeline: PipelineBase
   :param X: If importance_method is permutation, the holdout X data to compute importance with. Ignored otherwise.
   :type X: pd.DataFrame
   :param y: The holdout y data, used to obtain the name of the target class. If importance_method is permutation, used to compute importance with.
   :type y: pd.Series
   :param importance_method: The method of determining feature importance. One of ["permutation", "feature"]. Defaults to "permutation".
   :type importance_method: str
   :param max_features: The maximum number of influential features to include in an explanation. This does not affect the number of detrimental features reported. Defaults to 5.
   :type max_features: int
   :param min_importance_threshold: The minimum percent of total importance a single feature can have to be considered important. Defaults to 0.05.
   :type min_importance_threshold: float
   :param objective: If importance_method is permutation, the objective to compute importance with. Ignored otherwise, defaults to "auto".
   :type objective: str, ObjectiveBase

   :raises ValueError: if any arguments passed in are invalid or the pipeline is not fitted.