scikit-learn · jeremiedbb · Jul 29, 2025 · Jul 31, 2025 · Jul 31, 2025 · Jul 31, 2025
diff --git a/build_tools/circle/doc_environment.yml b/build_tools/circle/doc_environment.yml
@@ -39,6 +39,7 @@ dependencies:
   - towncrier
   - jupyterlite-sphinx
   - jupyterlite-pyodide-kernel
+  - ipympl
   - pip
   - pip:
     - sphinxcontrib-sass
diff --git a/build_tools/circle/doc_linux-64_conda.lock b/build_tools/circle/doc_linux-64_conda.lock
diff --git a/build_tools/update_environments_and_lock_files.py b/build_tools/update_environments_and_lock_files.py
@@ -378,6 +378,7 @@ def remove_from(alist, to_remove):
             "towncrier",
             "jupyterlite-sphinx",
             "jupyterlite-pyodide-kernel",
+            "ipympl",
         ],
         "pip_dependencies": [
             "sphinxcontrib-sass",

diff --git a/doc/conf.py b/doc/conf.py
@@ -745,6 +745,7 @@ def reset_sklearn_config(gallery_conf, fname):
     "plot_gallery": "True",
     "recommender": {"enable": True, "n_examples": 4, "min_df": 12},
     "reset_modules": ("matplotlib", "seaborn", reset_sklearn_config),
+    "first_notebook_cell": "%matplotlib widget",
 }
 if with_jupyterlite:
     sphinx_gallery_conf["jupyterlite"] = {

diff --git a/sklearn/metrics/_plot/precision_recall_curve.py b/sklearn/metrics/_plot/precision_recall_curve.py
@@ -8,11 +8,12 @@
     _BinaryClassifierCurveDisplayMixin,
     _deprecate_y_pred_parameter,
     _despine,
+    _LineTooltipMixin,
     _validate_style_kwargs,
 )
 
 
-class PrecisionRecallDisplay(_BinaryClassifierCurveDisplayMixin):
+class PrecisionRecallDisplay(_BinaryClassifierCurveDisplayMixin, _LineTooltipMixin):
     """Precision Recall visualization.
 
     It is recommended to use
@@ -34,6 +35,13 @@ class PrecisionRecallDisplay(_BinaryClassifierCurveDisplayMixin):
     recall : ndarray
         Recall values.
 
+    thresholds : ndarray or list of ndarrays, default=None
+        The thresholds at which the fpr and tpr have been computed. Each ndarray should
+        contain values for a single curve. If plotting multiple curves, list should be
+        of same length as `fpr` and `tpr`.
+        Only used to display the threshold values along the curve as a tooltip. If None,
+        only the fpr and tpr values are displayed.
+
     average_precision : float, default=None
         Average precision. If None, the average precision is not shown.
 
@@ -117,6 +125,7 @@ def __init__(
         precision,
         recall,
         *,
+        thresholds=None,
         average_precision=None,
         estimator_name=None,
         pos_label=None,
@@ -125,6 +134,7 @@ def __init__(
         self.estimator_name = estimator_name
         self.precision = precision
         self.recall = recall
+        self.thresholds = thresholds
         self.average_precision = average_precision
         self.pos_label = pos_label
         self.prevalence_pos_label = prevalence_pos_label
@@ -206,6 +216,10 @@ def plot(
 
         (self.line_,) = self.ax_.plot(self.recall, self.precision, **line_kwargs)
 
+        self._add_line_tooltip(
+            x_label="R", y_label="P", t_label="threshold", t_vals=[self.thresholds]
+        )
+
         info_pos_label = (
             f" (Positive label: {self.pos_label})" if self.pos_label is not None else ""
         )
@@ -537,7 +551,7 @@ def from_predictions(
             y_true, y_score, sample_weight=sample_weight, pos_label=pos_label, name=name
         )
 
-        precision, recall, _ = precision_recall_curve(
+        precision, recall, thresholds = precision_recall_curve(
             y_true,
             y_score,
             pos_label=pos_label,
@@ -554,6 +568,7 @@ def from_predictions(
         viz = cls(
             precision=precision,
             recall=recall,
+            thresholds=thresholds,
             average_precision=average_precision,
             estimator_name=name,
             pos_label=pos_label,

diff --git a/sklearn/metrics/_plot/roc_curve.py b/sklearn/metrics/_plot/roc_curve.py
@@ -13,12 +13,13 @@
     _deprecate_estimator_name,
     _deprecate_y_pred_parameter,
     _despine,
+    _LineTooltipMixin,
     _validate_style_kwargs,
 )
 from sklearn.utils._response import _get_response_values_binary
 
 
-class RocCurveDisplay(_BinaryClassifierCurveDisplayMixin):
+class RocCurveDisplay(_BinaryClassifierCurveDisplayMixin, _LineTooltipMixin):
     """ROC Curve visualization.
 
     It is recommended to use
@@ -49,6 +50,13 @@ class RocCurveDisplay(_BinaryClassifierCurveDisplayMixin):
         .. versionchanged:: 1.7
             Now accepts a list for plotting multiple curves.
 
+    thresholds : ndarray or list of ndarrays, default=None
+        The thresholds at which the fpr and tpr have been computed. Each ndarray should
+        contain values for a single curve. If plotting multiple curves, list should be
+        of same length as `fpr` and `tpr`.
+        Only used to display the threshold values along the curve as a tooltip. If None,
+        only the fpr and tpr values are displayed.
+
     roc_auc : float or list of floats, default=None
         Area under ROC curve, used for labeling each curve in the legend.
         If plotting multiple curves, should be a list of the same length as `fpr`
@@ -132,13 +140,15 @@ def __init__(
         *,
         fpr,
         tpr,
+        thresholds=None,
         roc_auc=None,
         name=None,
         pos_label=None,
         estimator_name="deprecated",
     ):
         self.fpr = fpr
         self.tpr = tpr
+        self.thresholds = thresholds
         self.roc_auc = roc_auc
         self.name = _deprecate_estimator_name(estimator_name, name, "1.7")
         self.pos_label = pos_label
@@ -148,6 +158,7 @@ def _validate_plot_params(self, *, ax, name):
 
         fpr = _convert_to_list_leaving_none(self.fpr)
         tpr = _convert_to_list_leaving_none(self.tpr)
+        thresholds = _convert_to_list_leaving_none(self.thresholds)
         roc_auc = _convert_to_list_leaving_none(self.roc_auc)
         name = _convert_to_list_leaving_none(name)
 
@@ -159,7 +170,7 @@ def _validate_plot_params(self, *, ax, name):
             optional=optional,
             class_name="RocCurveDisplay",
         )
-        return fpr, tpr, roc_auc, name
+        return fpr, tpr, thresholds, roc_auc, name
 
     def plot(
         self,
@@ -233,7 +244,9 @@ def plot(
         display : :class:`~sklearn.metrics.RocCurveDisplay`
             Object that stores computed values.
         """
-        fpr, tpr, roc_auc, name = self._validate_plot_params(ax=ax, name=name)
+        fpr, tpr, thresholds, roc_auc, name = self._validate_plot_params(
+            ax=ax, name=name
+        )
         n_curves = len(fpr)
         if not isinstance(curve_kwargs, list) and n_curves > 1:
             if roc_auc:
@@ -273,6 +286,10 @@ def plot(
         if len(self.line_) == 1:
             self.line_ = self.line_[0]
 
+        self._add_line_tooltip(
+            x_label="FPR", y_label="TPR", t_label="threshold", t_vals=thresholds
+        )
+
         info_pos_label = (
             f" (Positive label: {self.pos_label})" if self.pos_label is not None else ""
         )
@@ -580,7 +597,7 @@ def from_predictions(
             y_true, y_score, sample_weight=sample_weight, pos_label=pos_label, name=name
         )
 
-        fpr, tpr, _ = roc_curve(
+        fpr, tpr, thresholds = roc_curve(
             y_true,
             y_score,
             pos_label=pos_label,
@@ -592,6 +609,7 @@ def from_predictions(
         viz = cls(
             fpr=fpr,
             tpr=tpr,
+            thresholds=thresholds,
             roc_auc=roc_auc,
             name=name,
             pos_label=pos_label_validated,
@@ -731,7 +749,7 @@ def from_cv_results(
             pos_label=pos_label,
         )
 
-        fpr_folds, tpr_folds, auc_folds = [], [], []
+        fpr_folds, tpr_folds, thresholds_folds, auc_folds = [], [], [], []
         for estimator, test_indices in zip(
             cv_results["estimator"], cv_results["indices"]["test"]
         ):
@@ -747,7 +765,7 @@ def from_cv_results(
                 if sample_weight is None
                 else _safe_indexing(sample_weight, test_indices)
             )
-            fpr, tpr, _ = roc_curve(
+            fpr, tpr, thresholds = roc_curve(
                 y_true,
                 y_pred,
                 pos_label=pos_label_,
@@ -758,11 +776,13 @@ def from_cv_results(
 
             fpr_folds.append(fpr)
             tpr_folds.append(tpr)
+            thresholds_folds.append(thresholds)
             auc_folds.append(roc_auc)
 
         viz = cls(
             fpr=fpr_folds,
             tpr=tpr_folds,
+            thresholds=thresholds_folds,
             roc_auc=auc_folds,
             name=name,
             pos_label=pos_label_,

diff --git a/sklearn/metrics/_plot/tests/test_roc_curve_display.py b/sklearn/metrics/_plot/tests/test_roc_curve_display.py
@@ -16,6 +16,7 @@
 from sklearn.preprocessing import StandardScaler
 from sklearn.utils import _safe_indexing, shuffle
 from sklearn.utils._response import _get_response_values_binary
+from sklearn.utils.tests.test_plotting import _simulate_mouse_event
 
 
 @pytest.fixture(scope="module")
@@ -201,25 +202,29 @@ def test_validate_plot_params(pyplot):
     """Check `_validate_plot_params` returns the correct variables."""
     fpr = np.array([0, 0.5, 1])
     tpr = [np.array([0, 0.5, 1])]
+    thresholds = np.array([1, 0.5, 0])
     roc_auc = None
     name = "test_curve"
 
     # Initialize display with test inputs
     display = RocCurveDisplay(
         fpr=fpr,
         tpr=tpr,
+        thresholds=thresholds,
         roc_auc=roc_auc,
         name=name,
         pos_label=None,
     )
-    fpr_out, tpr_out, roc_auc_out, name_out = display._validate_plot_params(
-        ax=None, name=None
+    fpr_out, tpr_out, thresholds_out, roc_auc_out, name_out = (
+        display._validate_plot_params(ax=None, name=None)
     )
 
     assert isinstance(fpr_out, list)
     assert isinstance(tpr_out, list)
+    assert isinstance(thresholds_out, list)
     assert len(fpr_out) == 1
     assert len(tpr_out) == 1
+    assert len(thresholds_out) == 1
     assert roc_auc_out is None
     assert name_out == ["test_curve"]
 
@@ -980,3 +985,21 @@ def test_plot_roc_curve_despine(pyplot, data_binary, despine, constructor_name):
     if despine:
         for s in ["bottom", "left"]:
             assert display.ax_.spines[s].get_bounds() == (0, 1)
+
+
+def test_roc_curve_display_line_tooltip(pyplot, data_binary):
+    """Test the line tooltip on a Roc curve."""
+    X, y = data_binary
+
+    lr = LogisticRegression().fit(X, y)
+    display = RocCurveDisplay.from_estimator(lr, X, y)
+
+    assert hasattr(display, "line_tooltip_")
+
+    # simulate a mouse event on the line
+    x, y = display.fpr[10], display.tpr[10]
+    _simulate_mouse_event(display, x, y)
+
+    assert display.line_tooltip_.get_visible()
+    text = display.line_tooltip_.get_text()
+    assert all(name in text for name in ("FPR", "TPR", "threshold"))