scikit-learn · OmarManzoor · Apr 16, 2024 · Apr 15, 2024 · Apr 15, 2024 · Apr 15, 2024
diff --git a/doc/whats_new/v1.5.rst b/doc/whats_new/v1.5.rst
@@ -127,6 +127,13 @@ Changelog
    :pr:`123456` by :user:`Joe Bloggs <joeongithub>`.
    where 123455 is the *pull request* number, not the issue number.

+:mod:`sklearn.calibration`
+..........................
+
+- |Fix| Fixed a regression in :class:`calibration.CalibratedClassifierCV` where
+  an error was wrongly raised with string targets. 
+  :pr:`28843` by :user:`Jérémie du Boisberranger <jeremiedbb>`.
+
 :mod:`sklearn.cluster`
 ......................


diff --git a/sklearn/calibration.py b/sklearn/calibration.py
@@ -388,9 +388,7 @@ def fit(self, X, y, sample_weight=None, **fit_params):
                n_folds = self.cv.n_splits
            else:
                n_folds = None
-            if n_folds and np.any(
-                [np.sum(y == class_) < n_folds for class_ in self.classes_]
-            ):
+            if n_folds and np.any(np.unique(y, return_counts=True)[1] < n_folds):
                raise ValueError(
                    f"Requesting {n_folds}-fold "
                    "cross-validation but provided less than "

diff --git a/sklearn/tests/test_calibration.py b/sklearn/tests/test_calibration.py
@@ -1088,3 +1088,14 @@ def predict_proba(self, X):
    calibrator = CalibratedClassifierCV(model)
    # Does not raise an error
    calibrator.fit(*data)
+
+
+def test_error_less_class_samples_than_folds():
+    """Check that CalibratedClassifierCV works with string targets.
+
+    non-regression test for issue #28841.
+    """
+    X = np.random.normal(size=(20, 3))
+    y = ["a"] * 10 + ["b"] * 10
+
+    CalibratedClassifierCV(cv=3).fit(X, y)