holisticai.explainability.metrics.surrogate_features_stability

holisticai.explainability.metrics.surrogate_features_stability(X, y_pred, surrogate, num_bootstraps=5)

Calculate the stability of features used in a surrogate model. The metric measures the similarity of features used in the surrogate model across different bootstraps.

Parameters

XAny

The input data.

y_predAny

The predicted target values of the original model.

surrogateSurrogate

The surrogate model.

num_bootstrapsint

The number of bootstraps to use.

Returns

float

The stability of features used in a surrogate model.

Examples

>>> import numpy as np
>>> from holisticai.explainability.metrics.surrogate import (
...     surrogate_features_stability,
... )
>>> from sklearn.ensemble import RandomForestClassifier
>>> from sklearn.datasets import load_iris
>>> X, y = load_iris(return_X_y=True)
>>> model = RandomForestClassifier()
>>> model.fit(X, y)
>>> y_pred = model.predict(X)
>>> surrogate = RandomForestClassifier()
>>> surrogate.fit(X, y_pred)
>>> surrogate_features_stability(X, y_pred, surrogate)