Skip to content

BaseModelExplainer

Overview

A model_explainer is a component that takes is focused on model interpretability. It should be able to generated things like feature importance for models and explanations for predictions from a model.

Attributes

BaseModelExplainer contains no default attributes.

Configuration

BaseModelExplainer contains no required components.

Interface

The following methods are part of BaseModelExplainer and should be implemented in any class that inherits from this base class:

get_feature_importance

Generates feature importance for a given model. 

def get_feature_importance(self, data, model, *args, **kwargs) -> tuple(Any, Any)

Arguments:

  • data (dict): A dictionary of train/test data.
  • model (object): The model to use.

Returns:

  • explanations_train (Any): Explanations for the training dataset
  • explanations_test (Any): Explanations for the test dataset

get_explanations

Generates individual explanations for each row of the given dataset. 

def get_explanations(self, data, model, *args, **kwargs) -> Any

Arguments:

  • data (object): The data to generate explanations for. Likely a pandas.DataFrame.
  • model (object): The model to use.

Returns:

  • explanations (Any): Explanations for the model on the given data.