scikit-learn · jeremiedbb · May 7, 2025 · Mar 26, 2025 · Mar 30, 2025 · Apr 21, 2025
diff --git a/doc/whats_new/upcoming_changes/sklearn.feature_selection/31107.fix.rst b/doc/whats_new/upcoming_changes/sklearn.feature_selection/31107.fix.rst
@@ -0,0 +1,4 @@
+- :class:`feature_selection.SelectFromModel` now correctly works when the estimator
+  is an instance of :class:`linear_model.ElasticNetCV` with its `l1_ratio` parameter
+  being an array-like.
+  By :user:`Vasco Pereira <vasco-s-pereira>`.
diff --git a/sklearn/feature_selection/_from_model.py b/sklearn/feature_selection/_from_model.py
@@ -35,11 +35,18 @@ def _calculate_threshold(estimator, importances, threshold):
         est_name = estimator.__class__.__name__
         is_l1_penalized = hasattr(estimator, "penalty") and estimator.penalty == "l1"
         is_lasso = "Lasso" in est_name
-        is_elasticnet_l1_penalized = "ElasticNet" in est_name and (
-            (hasattr(estimator, "l1_ratio_") and np.isclose(estimator.l1_ratio_, 1.0))
-            or (hasattr(estimator, "l1_ratio") and np.isclose(estimator.l1_ratio, 1.0))
+        is_elasticnet_l1_penalized = est_name == "ElasticNet" and (
+            hasattr(estimator, "l1_ratio") and np.isclose(estimator.l1_ratio, 1.0)
         )
-        if is_l1_penalized or is_lasso or is_elasticnet_l1_penalized:
+        is_elasticnetcv_l1_penalized = est_name == "ElasticNetCV" and (
+            hasattr(estimator, "l1_ratio_") and np.isclose(estimator.l1_ratio_, 1.0)
+        )
+        if (
+            is_l1_penalized
+            or is_lasso
+            or is_elasticnet_l1_penalized
+            or is_elasticnetcv_l1_penalized
+        ):
             # the natural default threshold is 0 when l1 penalty was used
             threshold = 1e-5
         else:

diff --git a/sklearn/feature_selection/tests/test_from_model.py b/sklearn/feature_selection/tests/test_from_model.py
@@ -8,7 +8,7 @@
 from sklearn import datasets
 from sklearn.base import BaseEstimator
 from sklearn.cross_decomposition import CCA, PLSCanonical, PLSRegression
-from sklearn.datasets import make_friedman1
+from sklearn.datasets import make_friedman1, make_regression
 from sklearn.decomposition import PCA
 from sklearn.ensemble import HistGradientBoostingClassifier, RandomForestClassifier
 from sklearn.exceptions import NotFittedError
@@ -489,6 +489,21 @@ def test_prefit_max_features():
         model.transform(data)
 
 
+def test_get_feature_names_out_elasticnetcv():
+    """Check if ElasticNetCV works with a list of floats.
+
+    Non-regression test for #30936."""
+    X, y = make_regression(n_features=5, n_informative=3, random_state=0)
+    estimator = ElasticNetCV(l1_ratio=[0.25, 0.5, 0.75], random_state=0)
+    selector = SelectFromModel(estimator=estimator)
+    selector.fit(X, y)
+
+    names_out = selector.get_feature_names_out()
+    mask = selector.get_support()
+    expected = np.array([f"x{i}" for i in range(X.shape[1])])[mask]
+    assert_array_equal(names_out, expected)
+
+
 def test_prefit_get_feature_names_out():
     """Check the interaction between prefit and the feature names."""
     clf = RandomForestClassifier(n_estimators=2, random_state=0)