add

EstimatorReport.metrics.add(metric, *, name=None, greater_is_better=True, position='first', **kwargs)

Add a custom metric to be included in summarize() by default.

Parameters:

metricstr, sklearn scorer, or callable

The metric to add.

• If a string, it will be run through sklearn.metrics.get_scorer(). Metrics that require a neg_ prefix (e.g. "neg_mean_squared_error") may also be passed without it (e.g. "mean_squared_error"); the alias is resolved automatically.

• If a callable, it must have the signature (estimator, X, y_true, **kw) -> float. It may also return a dict mapping class labels to floats (e.g. {0: 0.9, 1: 0.85}), in which case summarize() will show one row per class label under the metric name. If your metric has the form (y_true, y_pred, **kw) -> float, see sklearn.metrics.make_scorer() to convert it to a scorer.

namestr, optional

Custom name for the metric. If not provided, the name is inferred from the metric (e.g. the function’s __name__).

greater_is_betterbool, default=True

Whether higher values are better (only for callables).

position{“first”, “last”}, default=”first”

Where to place the metric in default summarize() ordering. "first" inserts at the front; repeated "first" adds stack newest-first. "last" appends at the end.

**kwargsAny

Default keyword arguments passed to the score function at call time. Only used when metric is a plain callable.

Examples

>>> from sklearn.datasets import load_breast_cancer
>>> from sklearn.linear_model import LogisticRegression
>>> from sklearn.metrics import make_scorer, mean_absolute_error
>>> from skore import evaluate
>>> X, y = load_breast_cancer(return_X_y=True)
>>> classifier = LogisticRegression(max_iter=10_000)
>>> report = evaluate(classifier, X, y, splitter=0.2, pos_label=1)
>>> report.metrics.add(
...     make_scorer(mean_absolute_error, response_method="predict")
... )
>>> report.metrics.summarize().frame()
                    LogisticRegression
Metric
                                   ...
Mean Absolute Error                ...