scikit-learn · glemaitre · Jan 14, 2023 · Dec 21, 2022 · Dec 22, 2022 · Dec 22, 2022
diff --git a/doc/whats_new/v1.3.rst b/doc/whats_new/v1.3.rst
@@ -142,6 +142,12 @@ Changelog
   `feature_union["scalar"]`) to access transformers by name. :pr:`25093` by
   `Thomas Fan`_.
 
+- |Feature| :class:`pipeline.FeatureUnion` can now access the
+  `feature_names_in_` attribute if the `X` value seen during `.fit` has a
+  `columns` attribute and all columns are strings. e.g. when `X` is a
+  `pandas.DataFrame`
+  :pr:`25220` by :user:`Ian Thompson <it176131>`.
+
 :mod:`sklearn.preprocessing`
 ............................
 

diff --git a/sklearn/pipeline.py b/sklearn/pipeline.py
@@ -995,6 +995,12 @@ class FeatureUnion(TransformerMixin, _BaseComposition):
 
         .. versionadded:: 0.24
 
+    feature_names_in_ : ndarray of shape (`n_features_in_`,)
+        Names of features seen during :term:`fit`. Defined only when
+        `X` has feature names that are all strings.
+
+        .. versionadded:: 1.3
+
     See Also
     --------
     make_union : Convenience function for simplified feature union
@@ -1297,6 +1303,12 @@ def n_features_in_(self):
         # X is passed to all transformers so we just delegate to the first one
         return self.transformer_list[0][1].n_features_in_
 
+    @property
+    def feature_names_in_(self):
+        """Names of features seen during :term:`fit`."""
+        # X is passed to all transformers -- delegate to the first one
+        return self.transformer_list[0][1].feature_names_in_
+
     def __sklearn_is_fitted__(self):
         # Delegate whether feature union was fitted
         for _, transformer, _ in self._iter():

diff --git a/sklearn/tests/test_pipeline.py b/sklearn/tests/test_pipeline.py
@@ -1647,3 +1647,35 @@ def test_feature_union_getitem_error(key):
     msg = "Only string keys are supported"
     with pytest.raises(KeyError, match=msg):
         union[key]
+
+
+def test_feature_union_feature_names_in_():
+    """Ensure feature union has `.feature_names_in_` attribute if `X` has a
+    `columns` attribute.
+
+    Test for #24754.
+    """
+    pytest.importorskip("pandas")
+
+    X, _ = load_iris(as_frame=True, return_X_y=True)
+
+    # FeatureUnion should have the feature_names_in_ attribute if the
+    # first transformer also has it
+    scaler = StandardScaler()
+    scaler.fit(X)
+    union = FeatureUnion([("scale", scaler)])
+    assert hasattr(union, "feature_names_in_")
+    assert_array_equal(X.columns, union.feature_names_in_)
+    assert_array_equal(scaler.feature_names_in_, union.feature_names_in_)
+
+    # fit with pandas.DataFrame
+    union = FeatureUnion([("pass", "passthrough")])
+    union.fit(X)
+    assert hasattr(union, "feature_names_in_")
+    assert_array_equal(X.columns, union.feature_names_in_)
+
+    # fit with numpy array
+    X_array = X.to_numpy()
+    union = FeatureUnion([("pass", "passthrough")])
+    union.fit(X_array)
+    assert not hasattr(union, "feature_names_in_")