feat(ComparisonReport[CVReport]): Add ROC curve

[auguste-probabl]

Demo:

from sklearn.datasets import make_classification
from sklearn.linear_model import LogisticRegression
from skore import ComparisonReport, CrossValidationReport
X, y = make_classification(class_sep=0.1, random_state=42)
estimator_1 = LogisticRegression()
estimator_2 = LogisticRegression(C=2)  # Different regularization
report_1 = CrossValidationReport(estimator_1, X, y)
report_2 = CrossValidationReport(estimator_2, X, y)
report = ComparisonReport({"model1": report_1, "model2": report_2})
report.metrics.roc().plot()

from sklearn.datasets import make_classification
from sklearn.linear_model import LogisticRegression
from skore import ComparisonReport, CrossValidationReport
X, y = make_classification(
  class_sep=0.1, n_classes=3, n_clusters_per_class=1, random_state=42
)
estimator_1 = LogisticRegression()
estimator_2 = LogisticRegression(C=2)  # Different regularization
report_1 = CrossValidationReport(estimator_1, X, y)
report_2 = CrossValidationReport(estimator_2, X, y)
report = ComparisonReport({"model1": report_1, "model2": report_2})
display = report.metrics.roc()
display.plot()

---

Depends on #1668, #1682

TODO:

• n_curves is not calculated correctly in the comparison-cv case (unclear in what way)
• Binary classification: legend of chance level is not added for some reason (fixed in fix(roc-curve): Show chance level in the legend #1682)
• Move new tests closer to existing tests
• Deal with multiclass + pos_label is not None case: pos_label has no effect

[github-actions]

Documentation preview @ 47cd9d4

[auguste-probabl]

Note that the legend placement for the multiclass case depends on the size of the plot: here is what happens if the window is small:

[github-actions]

Coverage Report for skore/

File	Stmts	Miss	Cover	Missing
venv/lib/python3.12/site-packages/skore
__init__.py	23	0	100%
_config.py	28	0	100%
exceptions.py	4	4	0%	4, 15, 19, 23
venv/lib/python3.12/site-packages/skore/project
__init__.py	2	0	100%
metadata.py	67	0	100%
project.py	43	0	100%
reports.py	11	0	100%
widget.py	138	5	96%	375–377, 447–448
venv/lib/python3.12/site-packages/skore/sklearn
__init__.py	6	0	100%
_base.py	169	14	91%	45, 58, 126, 129, 182, 185–186, 188–191, 224, 227–228
find_ml_task.py	61	0	100%
types.py	20	0	100%
venv/lib/python3.12/site-packages/skore/sklearn/_comparison
__init__.py	5	0	100%
metrics_accessor.py	205	3	98%	165, 329, 1283
report.py	95	0	100%
utils.py	58	0	100%
venv/lib/python3.12/site-packages/skore/sklearn/_cross_validation
__init__.py	5	0	100%
metrics_accessor.py	204	1	99%	321
report.py	108	0	100%
venv/lib/python3.12/site-packages/skore/sklearn/_estimator
__init__.py	7	0	100%
feature_importance_accessor.py	143	2	98%	216–217
metrics_accessor.py	367	8	97%	181, 183, 190, 281, 350, 354, 369, 404
report.py	155	0	100%
venv/lib/python3.12/site-packages/skore/sklearn/_plot
__init__.py	2	0	100%
base.py	5	0	100%
style.py	28	0	100%
utils.py	118	5	95%	50, 74–76, 80
venv/lib/python3.12/site-packages/skore/sklearn/_plot/metrics
__init__.py	5	0	100%
confusion_matrix.py	69	4	94%	90, 98, 120, 228
precision_recall_curve.py	174	2	98%	521, 524
prediction_error.py	160	0	100%
roc_curve.py	242	4	98%	380, 497, 598, 791
venv/lib/python3.12/site-packages/skore/sklearn/train_test_split
__init__.py	0	0	100%
train_test_split.py	49	0	100%
venv/lib/python3.12/site-packages/skore/sklearn/train_test_split/warning
__init__.py	8	0	100%
high_class_imbalance_too_few_examples_warning.py	17	1	94%	80
high_class_imbalance_warning.py	18	0	100%
random_state_unset_warning.py	10	0	100%
shuffle_true_warning.py	10	1	90%	46
stratify_is_set_warning.py	10	0	100%
time_based_column_warning.py	21	1	95%	73
train_test_split_warning.py	4	0	100%
venv/lib/python3.12/site-packages/skore/utils
__init__.py	6	2	66%	8, 13
_accessor.py	52	2	96%	67, 108
_environment.py	27	0	100%
_fixes.py	8	0	100%
_index.py	5	0	100%
_logger.py	22	4	81%	15–17, 19
_measure_time.py	10	0	100%
_parallel.py	38	3	92%	23, 33, 124
_patch.py	13	5	61%	21, 23–24, 35, 37
_progress_bar.py	45	0	100%
_show_versions.py	33	2	93%	65–66
_testing.py	12	0	100%
TOTAL	3145	73	97%

Tests	Skipped	Failures	Errors	Time
786	5 💤	0 ❌	0 🔥	59.970s ⏱️

[glemaitre]

It is a partial review. I still have to look in details the roc_curve file and the tests but I thought that it could already be useful to look at those.

[glemaitre]

OK, I check in depth the roc_curve file. I have a couple of nit pickings.

Otherwise, I think that we can add a couple of tests for the kwargs (error and behaviour) and checking the dataframe as well.

I need to figure out some details for the legend.

[thomass-dev]

[automated comment] Please update your PR with main, so that the pytest workflow status will be reported.

[glemaitre]

LGTM. We can go ahead and revisit the legend all at once.

[glemaitre]

Nice, two plots to go. @sylvaincom You could start to look at creating the .frame specifically for this display. This way we could start to check the API.