holisticai.explainability.metrics.importance_stability

holisticai.explainability.metrics.importance_stability(feature_importances: ndarray, aggregate=True)

Determine the stability of feature importance.

Parameters

feature_importances: np.array A matrix of shape (M, d), where M is the number of samples and d is the number of features. Each entry represents the importance of a feature in a sample.

Returns

stability: float The stability metric, bounded between 0 and 1.