aif360.sklearn.metrics.intersection

aif360.sklearn.metrics.intersection(func, y_true, y_pred=None, prot_attr=None, sample_weight=None, return_groups=False, **kwargs)[source]

Compute an arbitrary metric on all intersectional groups of the protected attributes provided.

Parameters:

  • func (function) – A metric function from sklearn.metrics or aif360.sklearn.metrics.

  • y_true (pandas.Series) – Outcome vector with protected attributes as index.

  • y_pred (array-like, optional) – Estimated outcomes.

  • prot_attr (array-like, keyword-only) – Protected attribute(s). If None, all protected attributes in y are used.

  • sample_weight (array-like, optional) – Sample weights passed through to func.

  • return_groups (bool, optional) – Return group names in addition to metric values. Names are tuples of protected attribute values.

  • **kwargs – Additional keyword args to be passed through to func.

Returns:

list – List of metric values for each intersectional group.

tuple:

Metric values and their corresponding group names.

  • vals (list) – List of metric values for each intersectional group

  • groups (numpy.ndarray) – Array of tuples containing unique intersectional groups derived from the provided protected attributes.

Examples

>>> X, y = fetch_german()
>>> v, k = intersection(base_rate, y, prot_attr=['sex', 'age'],
...                     return_groups=True, pos_label='good')
>>> dict(zip(k, v))
{('female', 'aged'): 0.697560975609756,
 ('female', 'young'): 0.5523809523809524,
 ('male', 'aged'): 0.7388429752066116,
 ('male', 'young'): 0.611764705882353}