scikit-learn
diff --git a/‎doc/whats_new/v1.3.rst
Lines changed: 7 additions & 0 deletions b/‎doc/whats_new/v1.3.rst
Lines changed: 7 additions & 0 deletions
diff --git a/‎sklearn/impute/_knn.py
Lines changed: 6 additions & 1 deletion b/‎sklearn/impute/_knn.py
Lines changed: 6 additions & 1 deletion
diff --git a/‎sklearn/impute/tests/test_common.py
Lines changed: 36 additions & 0 deletions b/‎sklearn/impute/tests/test_common.py
Lines changed: 36 additions & 0 deletions
@@ -12,6 +12,13 @@ Version 1.3.1
 Changelog
 ---------
 
+:mod:`sklearn.impute`
+.....................
+
+- |Fix| :class:`impute.KNNImputer` now correctly adds a missing indicator column in
+  ``transform`` when ``add_indicator`` is set to ``True`` and missing values are observed
+  during ``fit``. :pr:`26600` by :user:`Shreesha Kumar Bhat <Shreesha3112>`.
+
 :mod:`sklearn.neighbors`
 ........................
 
 
@@ -282,7 +282,12 @@ def transform(self, X):
                 Xc[:, ~valid_mask] = 0
             else:
                 Xc = X[:, valid_mask]
-            return Xc
+
+            # Even if there are no missing values in X, we still concatenate Xc
+            # with the missing value indicator matrix, X_indicator.
+            # This is to ensure that the output maintains consistency in terms
+            # of columns, regardless of whether missing values exist in X or not.
+            return super()._concatenate_indicator(Xc, X_indicator)
 
         row_missing_idx = np.flatnonzero(mask.any(axis=1))
 
 
@@ -181,3 +181,39 @@ def test_keep_empty_features(imputer, keep_empty_features):
             assert X_imputed.shape == X.shape
         else:
             assert X_imputed.shape == (X.shape[0], X.shape[1] - 1)
+
+
+@pytest.mark.parametrize("imputer", imputers(), ids=lambda x: x.__class__.__name__)
+@pytest.mark.parametrize("missing_value_test", [np.nan, 1])
+def test_imputation_adds_missing_indicator_if_add_indicator_is_true(
+    imputer, missing_value_test
+):
+    """Check that missing indicator always exists when add_indicator=True.
+
+    Non-regression test for gh-26590.
+    """
+    X_train = np.array([[0, np.NaN], [1, 2]])
+
+    # Test data where missing_value_test variable can be set to np.NaN or 1.
+    X_test = np.array([[0, missing_value_test], [1, 2]])
+
+    imputer.set_params(add_indicator=True)
+    imputer.fit(X_train)
+
+    X_test_imputed_with_indicator = imputer.transform(X_test)
+    assert X_test_imputed_with_indicator.shape == (2, 3)
+
+    imputer.set_params(add_indicator=False)
+    imputer.fit(X_train)
+    X_test_imputed_without_indicator = imputer.transform(X_test)
+    assert X_test_imputed_without_indicator.shape == (2, 2)
+
+    assert_allclose(
+        X_test_imputed_with_indicator[:, :-1], X_test_imputed_without_indicator
+    )
+    if np.isnan(missing_value_test):
+        expected_missing_indicator = [1, 0]
+    else:
+        expected_missing_indicator = [0, 0]
+
+    assert_allclose(X_test_imputed_with_indicator[:, -1], expected_missing_indicator)