holisticai.explainability.metrics.rank_consistency

holisticai.explainability.metrics.rank_consistency(local_importances_values: ndarray, weighted=False, aggregate=True)

Calculate the rank consistency of local feature importances through all the instances.

Parameters

local_importances_valuesnp.ndarray

A 2D array where each row represents the local feature importances for a specific instance.

weightedbool, optional

If True, the rank consistency will be weighted by the average local importances. Default is False.

aggregatebool, optional

If True, the function will return an aggregated consistency score. If False, it will return the consistency scores for each feature. Default is True.

Returns

float or np.ndarray

If aggregate is True, returns a single float representing the aggregated rank consistency. If aggregate is False, returns an array of consistency scores for each feature.