From b45a370b8fb203f2b389f49ba8b693850ca3e4bd Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Wed, 1 Nov 2023 13:20:09 +1100
Subject: [PATCH 01/15] add chain method

---
 doc/whats_new/v1.4.rst            |   6 ++
 sklearn/multioutput.py            | 130 ++++++++++++++++++------------
 sklearn/tests/test_multioutput.py |  74 +++++++++++++----
 3 files changed, 143 insertions(+), 67 deletions(-)

diff --git a/doc/whats_new/v1.4.rst b/doc/whats_new/v1.4.rst
index ed7469e61567c..ae0804f04db85 100644
--- a/doc/whats_new/v1.4.rst
+++ b/doc/whats_new/v1.4.rst
@@ -395,6 +395,12 @@ Changelog
   object in the parameter grid if it's an estimator. :pr:`26786` by `Adrin
   Jalali`_.
 
+:mod:`sklearn.multioutput`
+..........................
+
+- |Enhancement| `chain_method` parameter added to
+  `:class:``multioutput.ClassifierChain`. by :user:`Lucy Liu <lucyleeow>`
+
 :mod:`sklearn.neighbors`
 ........................
 
diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 5cf1eae96fd3b..ffe47cd9719b0 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -648,6 +648,42 @@ def _log_message(self, *, estimator_idx, n_estimators, processing_msg):
             return None
         return f"({estimator_idx} of {n_estimators}) {processing_msg}"
 
+    def _get_Y_output(self, X, *, output_method="predict"):
+        """Get predictions for each model in the chain."""
+        check_is_fitted(self)
+        X = self._validate_data(X, accept_sparse=True, reset=False)
+        Y_output_chain = np.zeros((X.shape[0], len(self.estimators_)))
+        Y_feature_chain = np.zeros((X.shape[0], len(self.estimators_)))
+
+        # Note `decision_function` output is 1d for binary targets
+        is_feature_multi = self.chain_method_ in ["predict_proba", "predict_log_proba"]
+        is_output_multi = output_method == "predict_proba"
+
+        for chain_idx, estimator in enumerate(self.estimators_):
+            previous_predictions = Y_feature_chain[:, :chain_idx]
+            if sp.issparse(X):
+                X_aug = sp.hstack((X, previous_predictions))
+            else:
+                X_aug = np.hstack((X, previous_predictions))
+
+            feature_func = getattr(estimator, self.chain_method_)
+            feature_predictions = feature_func(X_aug)
+            if is_feature_multi:
+                feature_predictions = feature_predictions[:, 1]
+            Y_feature_chain[:, chain_idx] = feature_predictions
+
+            output_func = getattr(estimator, output_method)
+            output_predictions = output_func(X_aug)
+            if is_output_multi:
+                output_predictions = output_predictions[:, 1]
+            Y_output_chain[:, chain_idx] = output_predictions
+
+        inv_order = np.empty_like(self.order_)
+        inv_order[self.order_] = np.arange(len(self.order_))
+        Y_output = Y_output_chain[:, inv_order]
+
+        return Y_output
+
     @abstractmethod
     def fit(self, X, Y, **fit_params):
         """Fit the model to data matrix X and targets Y.
@@ -710,6 +746,15 @@ def fit(self, X, Y, **fit_params):
         else:
             routed_params = Bunch(estimator=Bunch(fit=fit_params))
 
+        try:
+            self.chain_method_ = self.chain_method
+        except AttributeError:
+            self.chain_method_ = "predict"
+
+        # Allow for different chain methhods for `ClassifierChain`;
+        # proba methods produce 2d output (decision_function 1d for binary targets)
+        multi_output = self.chain_method_ in ["predict_proba", "predict_log_proba"]
+
         for chain_idx, estimator in enumerate(self.estimators_):
             message = self._log_message(
                 estimator_idx=chain_idx + 1,
@@ -727,8 +772,10 @@ def fit(self, X, Y, **fit_params):
             if self.cv is not None and chain_idx < len(self.estimators_) - 1:
                 col_idx = X.shape[1] + chain_idx
                 cv_result = cross_val_predict(
-                    self.base_estimator, X_aug[:, :col_idx], y=y, cv=self.cv
+                    self.base_estimator, X_aug[:, :col_idx], y=y, cv=self.cv, method=self.chain_method_,
                 )
+                if multi_output:
+                    cv_result = cv_result[:, -1]
                 if sp.issparse(X_aug):
                     X_aug[:, col_idx] = np.expand_dims(cv_result, 1)
                 else:
@@ -749,25 +796,7 @@ def predict(self, X):
         Y_pred : array-like of shape (n_samples, n_classes)
             The predicted values.
         """
-        check_is_fitted(self)
-        X = self._validate_data(X, accept_sparse=True, reset=False)
-        Y_pred_chain = np.zeros((X.shape[0], len(self.estimators_)))
-        for chain_idx, estimator in enumerate(self.estimators_):
-            previous_predictions = Y_pred_chain[:, :chain_idx]
-            if sp.issparse(X):
-                if chain_idx == 0:
-                    X_aug = X
-                else:
-                    X_aug = sp.hstack((X, previous_predictions))
-            else:
-                X_aug = np.hstack((X, previous_predictions))
-            Y_pred_chain[:, chain_idx] = estimator.predict(X_aug)
-
-        inv_order = np.empty_like(self.order_)
-        inv_order[self.order_] = np.arange(len(self.order_))
-        Y_pred = Y_pred_chain[:, inv_order]
-
-        return Y_pred
+        return self._get_Y_output(X)
 
 
 class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):
@@ -818,6 +847,14 @@ class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):
         - :term:`CV splitter`,
         - An iterable yielding (train, test) splits as arrays of indices.
 
+    chain_method : {'predict', 'predict_proba', 'predict_log_proba', \
+        'decision_function'}, default='predict'
+
+        Prediction method to be used by estimators in the chain for
+        the 'prediction' features of previous estimators in the chain.
+
+        .. versionadded:: 1.4
+
     random_state : int, RandomState instance or None, optional (default=None)
         If ``order='random'``, determines random number generation for the
         chain order.
@@ -844,6 +881,9 @@ class labels for each estimator in the chain.
     order_ : list
         The order of labels in the classifier chain.
 
+    chain_method_ : str
+        Prediction method to be used by estimators in the chain.
+
     n_features_in_ : int
         Number of features seen during :term:`fit`. Only defined if the
         underlying `base_estimator` exposes such an attribute when fit.
@@ -891,6 +931,21 @@ class labels for each estimator in the chain.
            [0.0321..., 0.9935..., 0.0626...]])
     """
 
+    _parameter_constraints: dict = {
+        **_BaseChain._parameter_constraints,
+        "chain_method": [StrOptions({"predict", "predict_proba", "predict_log_proba", "decision_function"})],
+    }
+
+    def __init__(self, base_estimator, *, order=None, cv=None, chain_method='predict', random_state=None, verbose=False):
+        super().__init__(
+            base_estimator,
+            order=order,
+            cv=cv,
+            random_state=random_state,
+            verbose=verbose,
+        )
+        self.chain_method = chain_method
+
     @_fit_context(
         # ClassifierChain.base_estimator is not validated yet
         prefer_skip_nested_validation=False
@@ -941,22 +996,7 @@ def predict_proba(self, X):
         Y_prob : array-like of shape (n_samples, n_classes)
             The predicted probabilities.
         """
-        X = self._validate_data(X, accept_sparse=True, reset=False)
-        Y_prob_chain = np.zeros((X.shape[0], len(self.estimators_)))
-        Y_pred_chain = np.zeros((X.shape[0], len(self.estimators_)))
-        for chain_idx, estimator in enumerate(self.estimators_):
-            previous_predictions = Y_pred_chain[:, :chain_idx]
-            if sp.issparse(X):
-                X_aug = sp.hstack((X, previous_predictions))
-            else:
-                X_aug = np.hstack((X, previous_predictions))
-            Y_prob_chain[:, chain_idx] = estimator.predict_proba(X_aug)[:, 1]
-            Y_pred_chain[:, chain_idx] = estimator.predict(X_aug)
-        inv_order = np.empty_like(self.order_)
-        inv_order[self.order_] = np.arange(len(self.order_))
-        Y_prob = Y_prob_chain[:, inv_order]
-
-        return Y_prob
+        return self._get_Y_output(X, output_method="predict_proba")
 
     @_available_if_base_estimator_has("decision_function")
     def decision_function(self, X):
@@ -973,23 +1013,7 @@ def decision_function(self, X):
             Returns the decision function of the sample for each model
             in the chain.
         """
-        X = self._validate_data(X, accept_sparse=True, reset=False)
-        Y_decision_chain = np.zeros((X.shape[0], len(self.estimators_)))
-        Y_pred_chain = np.zeros((X.shape[0], len(self.estimators_)))
-        for chain_idx, estimator in enumerate(self.estimators_):
-            previous_predictions = Y_pred_chain[:, :chain_idx]
-            if sp.issparse(X):
-                X_aug = sp.hstack((X, previous_predictions))
-            else:
-                X_aug = np.hstack((X, previous_predictions))
-            Y_decision_chain[:, chain_idx] = estimator.decision_function(X_aug)
-            Y_pred_chain[:, chain_idx] = estimator.predict(X_aug)
-
-        inv_order = np.empty_like(self.order_)
-        inv_order[self.order_] = np.arange(len(self.order_))
-        Y_decision = Y_decision_chain[:, inv_order]
-
-        return Y_decision
+        return self._get_Y_output(X, output_method="decision_function")
 
     def get_metadata_routing(self):
         """Get metadata routing of this object.
diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index 493d0fc7dc8b5..58a3a90bf7888 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -492,10 +492,11 @@ def generate_multilabel_dataset_with_correlations():
     return X, Y_multi
 
 
-def test_classifier_chain_fit_and_predict_with_linear_svc():
+@pytest.mark.parametrize("chain_method", ["predict", "decision_function"])
+def test_classifier_chain_fit_and_predict_with_linear_svc(chain_method):
     # Fit classifier chain and verify predict performance using LinearSVC
     X, Y = generate_multilabel_dataset_with_correlations()
-    classifier_chain = ClassifierChain(LinearSVC(dual="auto"))
+    classifier_chain = ClassifierChain(LinearSVC(dual="auto"), chain_method=chain_method)
     classifier_chain.fit(X, Y)
 
     Y_pred = classifier_chain.predict(X)
@@ -548,23 +549,35 @@ def test_classifier_chain_vs_independent_models():
     )
 
 
-def test_base_chain_fit_and_predict():
-    # Fit base chain and verify predict performance
+@pytest.mark.parametrize("chain_method", ["predict", "predict_proba", "predict_log_proba", "decision_function"])
+def test_classifier_chain_fit_and_predict(chain_method):
+    # Fit classifier chain and verify predict performance
     X, Y = generate_multilabel_dataset_with_correlations()
-    chains = [RegressorChain(Ridge()), ClassifierChain(LogisticRegression())]
-    for chain in chains:
-        chain.fit(X, Y)
-        Y_pred = chain.predict(X)
-        assert Y_pred.shape == Y.shape
-        assert [c.coef_.size for c in chain.estimators_] == list(
-            range(X.shape[1], X.shape[1] + Y.shape[1])
-        )
+    chain = ClassifierChain(LogisticRegression(), chain_method=chain_method)
+    chain.fit(X, Y)
+    Y_pred = chain.predict(X)
+    assert Y_pred.shape == Y.shape
+    assert [c.coef_.size for c in chain.estimators_] == list(
+        range(X.shape[1], X.shape[1] + Y.shape[1])
+    )
 
-    Y_prob = chains[1].predict_proba(X)
+    Y_prob = chain.predict_proba(X)
     Y_binary = Y_prob >= 0.5
     assert_array_equal(Y_binary, Y_pred)
 
-    assert isinstance(chains[1], ClassifierMixin)
+    assert isinstance(chain, ClassifierMixin)
+
+
+def test_regressor_chain_fit_and_predict():
+    # Fit regressor chain and verify Y and estimator coefficients shape
+    X, Y = generate_multilabel_dataset_with_correlations()
+    chain = RegressorChain(Ridge())
+    chain.fit(X, Y)
+    Y_pred = chain.predict(X)
+    assert Y_pred.shape == Y.shape
+    assert [c.coef_.size for c in chain.estimators_] == list(
+        range(X.shape[1], X.shape[1] + Y.shape[1])
+    )
 
 
 @pytest.mark.parametrize("csr_container", CSR_CONTAINERS)
@@ -620,6 +633,39 @@ def test_base_chain_crossval_fit_and_predict():
             assert mean_squared_error(Y, Y_pred_cv) < 0.25
 
 
+@pytest.mark.parametrize(
+        "chain_type, chain_method",
+        [
+            ("classifier", "predict"),
+            ("classifier", "predict_proba"),
+            ("classifier", "predict_log_proba"),
+            ("classifier", "decision_function"),
+            ("regressor", ""),
+        ]
+)
+def test_base_chain_crossval_fit_and_predict(chain_type, chain_method):
+    # Fit chain with cross_val_predict and verify predict
+    # performance
+    X, Y = generate_multilabel_dataset_with_correlations()
+
+    if chain_type == "classifier":
+        chain = ClassifierChain(LogisticRegression(), chain_method=chain_method)
+    else:
+        chain = RegressorChain(Ridge())
+    chain.fit(X, Y)
+    chain_cv = clone(chain).set_params(cv=3)
+    chain_cv.fit(X, Y)
+    Y_pred_cv = chain_cv.predict(X)
+    Y_pred = chain.predict(X)
+
+    assert Y_pred_cv.shape == Y_pred.shape
+    assert not np.all(Y_pred == Y_pred_cv)
+    if isinstance(chain, ClassifierChain):
+        assert jaccard_score(Y, Y_pred_cv, average="samples") > 0.4
+    else:
+        assert mean_squared_error(Y, Y_pred_cv) < 0.25
+
+
 @pytest.mark.parametrize(
     "estimator",
     [

From e4753e93122d723c4995d526b5c4d6ab3707c360 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Wed, 1 Nov 2023 13:20:47 +1100
Subject: [PATCH 02/15] black

---
 sklearn/multioutput.py            | 23 ++++++++++++++++++++---
 sklearn/tests/test_multioutput.py | 25 +++++++++++++++----------
 2 files changed, 35 insertions(+), 13 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index ffe47cd9719b0..3a987b86f1a17 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -772,7 +772,11 @@ def fit(self, X, Y, **fit_params):
             if self.cv is not None and chain_idx < len(self.estimators_) - 1:
                 col_idx = X.shape[1] + chain_idx
                 cv_result = cross_val_predict(
-                    self.base_estimator, X_aug[:, :col_idx], y=y, cv=self.cv, method=self.chain_method_,
+                    self.base_estimator,
+                    X_aug[:, :col_idx],
+                    y=y,
+                    cv=self.cv,
+                    method=self.chain_method_,
                 )
                 if multi_output:
                     cv_result = cv_result[:, -1]
@@ -933,10 +937,23 @@ class labels for each estimator in the chain.
 
     _parameter_constraints: dict = {
         **_BaseChain._parameter_constraints,
-        "chain_method": [StrOptions({"predict", "predict_proba", "predict_log_proba", "decision_function"})],
+        "chain_method": [
+            StrOptions(
+                {"predict", "predict_proba", "predict_log_proba", "decision_function"}
+            )
+        ],
     }
 
-    def __init__(self, base_estimator, *, order=None, cv=None, chain_method='predict', random_state=None, verbose=False):
+    def __init__(
+        self,
+        base_estimator,
+        *,
+        order=None,
+        cv=None,
+        chain_method="predict",
+        random_state=None,
+        verbose=False,
+    ):
         super().__init__(
             base_estimator,
             order=order,
diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index 58a3a90bf7888..17f56a225fb18 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -496,7 +496,9 @@ def generate_multilabel_dataset_with_correlations():
 def test_classifier_chain_fit_and_predict_with_linear_svc(chain_method):
     # Fit classifier chain and verify predict performance using LinearSVC
     X, Y = generate_multilabel_dataset_with_correlations()
-    classifier_chain = ClassifierChain(LinearSVC(dual="auto"), chain_method=chain_method)
+    classifier_chain = ClassifierChain(
+        LinearSVC(dual="auto"), chain_method=chain_method
+    )
     classifier_chain.fit(X, Y)
 
     Y_pred = classifier_chain.predict(X)
@@ -549,7 +551,10 @@ def test_classifier_chain_vs_independent_models():
     )
 
 
-@pytest.mark.parametrize("chain_method", ["predict", "predict_proba", "predict_log_proba", "decision_function"])
+@pytest.mark.parametrize(
+    "chain_method",
+    ["predict", "predict_proba", "predict_log_proba", "decision_function"],
+)
 def test_classifier_chain_fit_and_predict(chain_method):
     # Fit classifier chain and verify predict performance
     X, Y = generate_multilabel_dataset_with_correlations()
@@ -634,14 +639,14 @@ def test_base_chain_crossval_fit_and_predict():
 
 
 @pytest.mark.parametrize(
-        "chain_type, chain_method",
-        [
-            ("classifier", "predict"),
-            ("classifier", "predict_proba"),
-            ("classifier", "predict_log_proba"),
-            ("classifier", "decision_function"),
-            ("regressor", ""),
-        ]
+    "chain_type, chain_method",
+    [
+        ("classifier", "predict"),
+        ("classifier", "predict_proba"),
+        ("classifier", "predict_log_proba"),
+        ("classifier", "decision_function"),
+        ("regressor", ""),
+    ],
 )
 def test_base_chain_crossval_fit_and_predict(chain_type, chain_method):
     # Fit chain with cross_val_predict and verify predict

From 347dcbbdca609037fa449d33ee1739e584bcd949 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Wed, 1 Nov 2023 13:28:54 +1100
Subject: [PATCH 03/15] whats new

---
 doc/whats_new/v1.4.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/whats_new/v1.4.rst b/doc/whats_new/v1.4.rst
index ae0804f04db85..1dc85aa854e60 100644
--- a/doc/whats_new/v1.4.rst
+++ b/doc/whats_new/v1.4.rst
@@ -399,7 +399,7 @@ Changelog
 ..........................
 
 - |Enhancement| `chain_method` parameter added to
-  `:class:``multioutput.ClassifierChain`. by :user:`Lucy Liu <lucyleeow>`
+  `:class:``multioutput.ClassifierChain`. :pr:`27700` by :user:`Lucy Liu <lucyleeow>`
 
 :mod:`sklearn.neighbors`
 ........................

From d5752bf08538ab5210794cd073294e48cf5c79cf Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Wed, 1 Nov 2023 13:29:54 +1100
Subject: [PATCH 04/15] rm duplicated test

---
 sklearn/tests/test_multioutput.py | 20 --------------------
 1 file changed, 20 deletions(-)

diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index 17f56a225fb18..d2422ea868009 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -618,26 +618,6 @@ def test_base_chain_random_order():
             assert_array_almost_equal(est1.coef_, est2.coef_)
 
 
-def test_base_chain_crossval_fit_and_predict():
-    # Fit chain with cross_val_predict and verify predict
-    # performance
-    X, Y = generate_multilabel_dataset_with_correlations()
-
-    for chain in [ClassifierChain(LogisticRegression()), RegressorChain(Ridge())]:
-        chain.fit(X, Y)
-        chain_cv = clone(chain).set_params(cv=3)
-        chain_cv.fit(X, Y)
-        Y_pred_cv = chain_cv.predict(X)
-        Y_pred = chain.predict(X)
-
-        assert Y_pred_cv.shape == Y_pred.shape
-        assert not np.all(Y_pred == Y_pred_cv)
-        if isinstance(chain, ClassifierChain):
-            assert jaccard_score(Y, Y_pred_cv, average="samples") > 0.4
-        else:
-            assert mean_squared_error(Y, Y_pred_cv) < 0.25
-
-
 @pytest.mark.parametrize(
     "chain_type, chain_method",
     [

From 4dde42c3a641111e3650d6bd36377b46010ff0ed Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Wed, 1 Nov 2023 14:59:50 +1100
Subject: [PATCH 05/15] fix, rm chain_method param from regressorchain

---
 sklearn/multioutput.py            | 34 ++++++++++++++++++-------------
 sklearn/tests/test_multioutput.py |  2 +-
 2 files changed, 21 insertions(+), 15 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 3a987b86f1a17..a137846ed68db 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -648,6 +648,14 @@ def _log_message(self, *, estimator_idx, n_estimators, processing_msg):
             return None
         return f"({estimator_idx} of {n_estimators}) {processing_msg}"
 
+    def _get_chain_method(self):
+        # Different chain methods allowed in `ClassifierChain` only
+        try:
+            chain_method = self.chain_method_
+        except AttributeError:
+            chain_method = "predict"
+        return chain_method
+
     def _get_Y_output(self, X, *, output_method="predict"):
         """Get predictions for each model in the chain."""
         check_is_fitted(self)
@@ -655,8 +663,9 @@ def _get_Y_output(self, X, *, output_method="predict"):
         Y_output_chain = np.zeros((X.shape[0], len(self.estimators_)))
         Y_feature_chain = np.zeros((X.shape[0], len(self.estimators_)))
 
-        # Note `decision_function` output is 1d for binary targets
-        is_feature_multi = self.chain_method_ in ["predict_proba", "predict_log_proba"]
+        chain_method = self._get_chain_method()
+        # proba methods produce 2d output (decision_function 1d for binary targets)
+        is_feature_multi = chain_method in ["predict_proba", "predict_log_proba"]
         is_output_multi = output_method == "predict_proba"
 
         for chain_idx, estimator in enumerate(self.estimators_):
@@ -666,7 +675,7 @@ def _get_Y_output(self, X, *, output_method="predict"):
             else:
                 X_aug = np.hstack((X, previous_predictions))
 
-            feature_func = getattr(estimator, self.chain_method_)
+            feature_func = getattr(estimator, chain_method)
             feature_predictions = feature_func(X_aug)
             if is_feature_multi:
                 feature_predictions = feature_predictions[:, 1]
@@ -746,14 +755,9 @@ def fit(self, X, Y, **fit_params):
         else:
             routed_params = Bunch(estimator=Bunch(fit=fit_params))
 
-        try:
-            self.chain_method_ = self.chain_method
-        except AttributeError:
-            self.chain_method_ = "predict"
-
-        # Allow for different chain methhods for `ClassifierChain`;
+        chain_method = self._get_chain_method()
         # proba methods produce 2d output (decision_function 1d for binary targets)
-        multi_output = self.chain_method_ in ["predict_proba", "predict_log_proba"]
+        is_multi = chain_method in ["predict_proba", "predict_log_proba"]
 
         for chain_idx, estimator in enumerate(self.estimators_):
             message = self._log_message(
@@ -776,10 +780,10 @@ def fit(self, X, Y, **fit_params):
                     X_aug[:, :col_idx],
                     y=y,
                     cv=self.cv,
-                    method=self.chain_method_,
+                    method=chain_method,
                 )
-                if multi_output:
-                    cv_result = cv_result[:, -1]
+                if is_multi:
+                    cv_result = cv_result[:, 1]
                 if sp.issparse(X_aug):
                     X_aug[:, col_idx] = np.expand_dims(cv_result, 1)
                 else:
@@ -886,7 +890,8 @@ class labels for each estimator in the chain.
         The order of labels in the classifier chain.
 
     chain_method_ : str
-        Prediction method to be used by estimators in the chain.
+        Prediction method to be used by estimators in the chain for the 'prediction'
+        features.
 
     n_features_in_ : int
         Number of features seen during :term:`fit`. Only defined if the
@@ -993,6 +998,7 @@ def fit(self, X, Y, **fit_params):
         """
         _raise_for_params(fit_params, self, "fit")
 
+        self.chain_method_ = self.chain_method
         super().fit(X, Y, **fit_params)
         self.classes_ = [
             estimator.classes_ for chain_idx, estimator in enumerate(self.estimators_)
diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index d2422ea868009..fc620139c53f1 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -497,7 +497,7 @@ def test_classifier_chain_fit_and_predict_with_linear_svc(chain_method):
     # Fit classifier chain and verify predict performance using LinearSVC
     X, Y = generate_multilabel_dataset_with_correlations()
     classifier_chain = ClassifierChain(
-        LinearSVC(dual="auto"), chain_method=chain_method
+        LinearSVC(dual="auto"), chain_method=chain_method,
     )
     classifier_chain.fit(X, Y)
 

From 8aab5d6579f8ff1d0714cb476abe461f9b3dafa2 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Wed, 1 Nov 2023 15:00:17 +1100
Subject: [PATCH 06/15] black

---
 sklearn/tests/test_multioutput.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index fc620139c53f1..491021c8a90bd 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -497,7 +497,8 @@ def test_classifier_chain_fit_and_predict_with_linear_svc(chain_method):
     # Fit classifier chain and verify predict performance using LinearSVC
     X, Y = generate_multilabel_dataset_with_correlations()
     classifier_chain = ClassifierChain(
-        LinearSVC(dual="auto"), chain_method=chain_method,
+        LinearSVC(dual="auto"),
+        chain_method=chain_method,
     )
     classifier_chain.fit(X, Y)
 

From 9e50488250a20b4d0d1753da94e29522d8814fd5 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Thu, 2 Nov 2023 16:02:16 +1100
Subject: [PATCH 07/15] use _get_response_values

---
 sklearn/multioutput.py | 58 ++++++++++++++++++++++--------------------
 1 file changed, 31 insertions(+), 27 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index a137846ed68db..523c0c509e031 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -33,6 +33,8 @@
 from .model_selection import cross_val_predict
 from .utils import Bunch, _print_elapsed_time, check_random_state
 from .utils._param_validation import HasMethods, StrOptions
+from .utils._response import _get_response_values
+from .utils.validation import _check_response_method
 from .utils.metadata_routing import (
     MetadataRouter,
     MethodMapping,
@@ -649,14 +651,14 @@ def _log_message(self, *, estimator_idx, n_estimators, processing_msg):
         return f"({estimator_idx} of {n_estimators}) {processing_msg}"
 
     def _get_chain_method(self):
-        # Different chain methods allowed in `ClassifierChain` only
         try:
-            chain_method = self.chain_method_
+            chain_method = self.chain_method
+        # `RegressorChain` does not have a `chain_method` parameter
         except AttributeError:
             chain_method = "predict"
-        return chain_method
+        return _check_response_method(self.base_estimator, chain_method)
 
-    def _get_Y_output(self, X, *, output_method="predict"):
+    def _get_predictions(self, X, *, output_method):
         """Get predictions for each model in the chain."""
         check_is_fitted(self)
         X = self._validate_data(X, accept_sparse=True, reset=False)
@@ -664,10 +666,6 @@ def _get_Y_output(self, X, *, output_method="predict"):
         Y_feature_chain = np.zeros((X.shape[0], len(self.estimators_)))
 
         chain_method = self._get_chain_method()
-        # proba methods produce 2d output (decision_function 1d for binary targets)
-        is_feature_multi = chain_method in ["predict_proba", "predict_log_proba"]
-        is_output_multi = output_method == "predict_proba"
-
         for chain_idx, estimator in enumerate(self.estimators_):
             previous_predictions = Y_feature_chain[:, :chain_idx]
             if sp.issparse(X):
@@ -675,16 +673,18 @@ def _get_Y_output(self, X, *, output_method="predict"):
             else:
                 X_aug = np.hstack((X, previous_predictions))
 
-            feature_func = getattr(estimator, chain_method)
-            feature_predictions = feature_func(X_aug)
-            if is_feature_multi:
-                feature_predictions = feature_predictions[:, 1]
+            feature_predictions, _ = _get_response_values(
+                estimator,
+                X_aug,
+                response_method=chain_method,
+            )
             Y_feature_chain[:, chain_idx] = feature_predictions
 
-            output_func = getattr(estimator, output_method)
-            output_predictions = output_func(X_aug)
-            if is_output_multi:
-                output_predictions = output_predictions[:, 1]
+            output_predictions, _ = _get_response_values(
+                estimator,
+                X_aug,
+                response_method=output_method,
+            )
             Y_output_chain[:, chain_idx] = output_predictions
 
         inv_order = np.empty_like(self.order_)
@@ -756,9 +756,6 @@ def fit(self, X, Y, **fit_params):
             routed_params = Bunch(estimator=Bunch(fit=fit_params))
 
         chain_method = self._get_chain_method()
-        # proba methods produce 2d output (decision_function 1d for binary targets)
-        is_multi = chain_method in ["predict_proba", "predict_log_proba"]
-
         for chain_idx, estimator in enumerate(self.estimators_):
             message = self._log_message(
                 estimator_idx=chain_idx + 1,
@@ -782,7 +779,7 @@ def fit(self, X, Y, **fit_params):
                     cv=self.cv,
                     method=chain_method,
                 )
-                if is_multi:
+                if cv_result.ndim > 1:
                     cv_result = cv_result[:, 1]
                 if sp.issparse(X_aug):
                     X_aug[:, col_idx] = np.expand_dims(cv_result, 1)
@@ -804,7 +801,7 @@ def predict(self, X):
         Y_pred : array-like of shape (n_samples, n_classes)
             The predicted values.
         """
-        return self._get_Y_output(X)
+        return self._get_predictions(X)
 
 
 class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):
@@ -856,11 +853,16 @@ class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):
         - An iterable yielding (train, test) splits as arrays of indices.
 
     chain_method : {'predict', 'predict_proba', 'predict_log_proba', \
-        'decision_function'}, default='predict'
+        'decision_function'} or list of such str's, default='predict'
 
         Prediction method to be used by estimators in the chain for
         the 'prediction' features of previous estimators in the chain.
 
+        - if `str`, name of the method;
+        - if a list of `str`, provides the method names in order of
+          preference. The method used corresponds to the first method in
+          the list that is implemented by `base_estimator`.
+
         .. versionadded:: 1.4
 
     random_state : int, RandomState instance or None, optional (default=None)
@@ -890,7 +892,7 @@ class labels for each estimator in the chain.
         The order of labels in the classifier chain.
 
     chain_method_ : str
-        Prediction method to be used by estimators in the chain for the 'prediction'
+        Prediction method used by estimators in the chain for the 'prediction'
         features.
 
     n_features_in_ : int
@@ -943,9 +945,11 @@ class labels for each estimator in the chain.
     _parameter_constraints: dict = {
         **_BaseChain._parameter_constraints,
         "chain_method": [
+            list,
+            tuple,
             StrOptions(
                 {"predict", "predict_proba", "predict_log_proba", "decision_function"}
-            )
+            ),
         ],
     }
 
@@ -998,11 +1002,11 @@ def fit(self, X, Y, **fit_params):
         """
         _raise_for_params(fit_params, self, "fit")
 
-        self.chain_method_ = self.chain_method
         super().fit(X, Y, **fit_params)
         self.classes_ = [
             estimator.classes_ for chain_idx, estimator in enumerate(self.estimators_)
         ]
+        self.chain_method_ = self._get_chain_method()
         return self
 
     @_available_if_base_estimator_has("predict_proba")
@@ -1019,7 +1023,7 @@ def predict_proba(self, X):
         Y_prob : array-like of shape (n_samples, n_classes)
             The predicted probabilities.
         """
-        return self._get_Y_output(X, output_method="predict_proba")
+        return self._get_predictions(X, output_method="predict_proba")
 
     @_available_if_base_estimator_has("decision_function")
     def decision_function(self, X):
@@ -1036,7 +1040,7 @@ def decision_function(self, X):
             Returns the decision function of the sample for each model
             in the chain.
         """
-        return self._get_Y_output(X, output_method="decision_function")
+        return self._get_predictions(X, output_method="decision_function")
 
     def get_metadata_routing(self):
         """Get metadata routing of this object.

From 7ca530e0da576996ca6e8d3b6c5c52ba30860356 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Thu, 2 Nov 2023 16:12:25 +1100
Subject: [PATCH 08/15] fix-give output_method

---
 sklearn/multioutput.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 14c8e825b1ff7..76d1a0da6d27c 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -656,7 +656,8 @@ def _get_chain_method(self):
         # `RegressorChain` does not have a `chain_method` parameter
         except AttributeError:
             chain_method = "predict"
-        return _check_response_method(self.base_estimator, chain_method)
+        method = _check_response_method(self.base_estimator, chain_method)
+        return method.__name__
 
     def _get_predictions(self, X, *, output_method):
         """Get predictions for each model in the chain."""
@@ -801,7 +802,7 @@ def predict(self, X):
         Y_pred : array-like of shape (n_samples, n_classes)
             The predicted values.
         """
-        return self._get_predictions(X)
+        return self._get_predictions(X, output_method="predict")
 
 
 class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):

From 6d1eb7d6a1c4f9239ea7deee3277d564f7161df4 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Fri, 3 Nov 2023 10:15:47 +1100
Subject: [PATCH 09/15] fix fit chain_method_ attr

---
 sklearn/multioutput.py | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 76d1a0da6d27c..912c59cced7bd 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -650,15 +650,6 @@ def _log_message(self, *, estimator_idx, n_estimators, processing_msg):
             return None
         return f"({estimator_idx} of {n_estimators}) {processing_msg}"
 
-    def _get_chain_method(self):
-        try:
-            chain_method = self.chain_method
-        # `RegressorChain` does not have a `chain_method` parameter
-        except AttributeError:
-            chain_method = "predict"
-        method = _check_response_method(self.base_estimator, chain_method)
-        return method.__name__
-
     def _get_predictions(self, X, *, output_method):
         """Get predictions for each model in the chain."""
         check_is_fitted(self)
@@ -666,7 +657,9 @@ def _get_predictions(self, X, *, output_method):
         Y_output_chain = np.zeros((X.shape[0], len(self.estimators_)))
         Y_feature_chain = np.zeros((X.shape[0], len(self.estimators_)))
 
-        chain_method = self._get_chain_method()
+        # `RegressorChain` does not have a `chain_method_` parameter so we
+        # default to "predict"
+        chain_method = getattr(self, "chain_method_", "predict")
         for chain_idx, estimator in enumerate(self.estimators_):
             previous_predictions = Y_feature_chain[:, :chain_idx]
             if sp.issparse(X):
@@ -756,7 +749,15 @@ def fit(self, X, Y, **fit_params):
         else:
             routed_params = Bunch(estimator=Bunch(fit=fit_params))
 
-        chain_method = self._get_chain_method()
+        if hasattr(self, "chain_method"):
+            chain_method = _check_response_method(
+                self.base_estimator, self.chain_method,
+            ).__name__
+            self.chain_method_ = chain_method
+        else:
+            # `RegressorChain` does not have a `chain_method` parameter
+            chain_method = "predict"
+
         for chain_idx, estimator in enumerate(self.estimators_):
             message = self._log_message(
                 estimator_idx=chain_idx + 1,
@@ -1005,7 +1006,6 @@ def fit(self, X, Y, **fit_params):
 
         super().fit(X, Y, **fit_params)
         self.classes_ = [estimator.classes_ for estimator in self.estimators_]
-        self.chain_method_ = self._get_chain_method()
         return self
 
     @_available_if_base_estimator_has("predict_proba")

From dbe7ef27e1867d93edd1b0668d91c66b28772cc4 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Fri, 3 Nov 2023 10:16:06 +1100
Subject: [PATCH 10/15] black

---
 sklearn/multioutput.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 912c59cced7bd..ca234ae054cb9 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -751,7 +751,8 @@ def fit(self, X, Y, **fit_params):
 
         if hasattr(self, "chain_method"):
             chain_method = _check_response_method(
-                self.base_estimator, self.chain_method,
+                self.base_estimator,
+                self.chain_method,
             ).__name__
             self.chain_method_ = chain_method
         else:

From 41414511514fadb95e6816546371cf4ff2edc00d Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Mon, 4 Dec 2023 21:18:31 +1100
Subject: [PATCH 11/15] lint

---
 sklearn/multioutput.py | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index ca234ae054cb9..b07e0d09ceceb 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -34,7 +34,6 @@
 from .utils import Bunch, _print_elapsed_time, check_random_state
 from .utils._param_validation import HasMethods, StrOptions
 from .utils._response import _get_response_values
-from .utils.validation import _check_response_method
 from .utils.metadata_routing import (
     MetadataRouter,
     MethodMapping,
@@ -45,7 +44,12 @@
 from .utils.metaestimators import available_if
 from .utils.multiclass import check_classification_targets
 from .utils.parallel import Parallel, delayed
-from .utils.validation import _check_method_params, check_is_fitted, has_fit_parameter
+from .utils.validation import (
+    _check_method_params,
+    _check_response_method,
+    check_is_fitted,
+    has_fit_parameter,
+)
 
 __all__ = [
     "MultiOutputRegressor",

From c554ac52bf6960133e299ea5d53147374f86a560 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Tue, 5 Dec 2023 17:11:00 +1100
Subject: [PATCH 12/15] review formatting

---
 doc/whats_new/v1.4.rst            |  4 ++--
 sklearn/multioutput.py            | 15 +++++----------
 sklearn/tests/test_multioutput.py | 12 ++++--------
 3 files changed, 11 insertions(+), 20 deletions(-)

diff --git a/doc/whats_new/v1.4.rst b/doc/whats_new/v1.4.rst
index 00b0be9917961..a46b43bd46e33 100644
--- a/doc/whats_new/v1.4.rst
+++ b/doc/whats_new/v1.4.rst
@@ -532,8 +532,8 @@ Changelog
 :mod:`sklearn.multioutput`
 ..........................
 
-- |Enhancement| `chain_method` parameter added to
-  `:class:``multioutput.ClassifierChain`. :pr:`27700` by :user:`Lucy Liu <lucyleeow>`
+- |Enhancement| `chain_method` parameter added to `:class:``multioutput.ClassifierChain`.
+  :pr:`27700` by :user:`Lucy Liu <lucyleeow>`.
 
 :mod:`sklearn.neighbors`
 ........................
diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index b07e0d09ceceb..528751011654a 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -672,16 +672,12 @@ def _get_predictions(self, X, *, output_method):
                 X_aug = np.hstack((X, previous_predictions))
 
             feature_predictions, _ = _get_response_values(
-                estimator,
-                X_aug,
-                response_method=chain_method,
+                estimator, X_aug, response_method=chain_method,
             )
             Y_feature_chain[:, chain_idx] = feature_predictions
 
             output_predictions, _ = _get_response_values(
-                estimator,
-                X_aug,
-                response_method=output_method,
+                estimator, X_aug, response_method=output_method,
             )
             Y_output_chain[:, chain_idx] = output_predictions
 
@@ -755,8 +751,7 @@ def fit(self, X, Y, **fit_params):
 
         if hasattr(self, "chain_method"):
             chain_method = _check_response_method(
-                self.base_estimator,
-                self.chain_method,
+                self.base_estimator, self.chain_method,
             ).__name__
             self.chain_method_ = chain_method
         else:
@@ -860,7 +855,7 @@ class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):
         - An iterable yielding (train, test) splits as arrays of indices.
 
     chain_method : {'predict', 'predict_proba', 'predict_log_proba', \
-        'decision_function'} or list of such str's, default='predict'
+            'decision_function'} or list of such str's, default='predict'
 
         Prediction method to be used by estimators in the chain for
         the 'prediction' features of previous estimators in the chain.
@@ -899,7 +894,7 @@ class labels for each estimator in the chain.
         The order of labels in the classifier chain.
 
     chain_method_ : str
-        Prediction method used by estimators in the chain for the 'prediction'
+        Prediction method used by estimators in the chain for the prediction
         features.
 
     n_features_in_ : int
diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index 491021c8a90bd..d16d29d460776 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -497,10 +497,8 @@ def test_classifier_chain_fit_and_predict_with_linear_svc(chain_method):
     # Fit classifier chain and verify predict performance using LinearSVC
     X, Y = generate_multilabel_dataset_with_correlations()
     classifier_chain = ClassifierChain(
-        LinearSVC(dual="auto"),
-        chain_method=chain_method,
-    )
-    classifier_chain.fit(X, Y)
+        LinearSVC(dual="auto"), chain_method=chain_method,
+    ).fit(X, Y)
 
     Y_pred = classifier_chain.predict(X)
     assert Y_pred.shape == Y.shape
@@ -518,12 +516,10 @@ def test_classifier_chain_fit_and_predict_with_sparse_data(csr_container):
     X, Y = generate_multilabel_dataset_with_correlations()
     X_sparse = csr_container(X)
 
-    classifier_chain = ClassifierChain(LogisticRegression())
-    classifier_chain.fit(X_sparse, Y)
+    classifier_chain = ClassifierChain(LogisticRegression()).fit(X_sparse, Y)
     Y_pred_sparse = classifier_chain.predict(X_sparse)
 
-    classifier_chain = ClassifierChain(LogisticRegression())
-    classifier_chain.fit(X, Y)
+    classifier_chain = ClassifierChain(LogisticRegression()).fit(X, Y)
     Y_pred_dense = classifier_chain.predict(X)
 
     assert_array_equal(Y_pred_sparse, Y_pred_dense)

From aa09f21f4043fe88a37bd0ebf3389c9bdcb2ea71 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Tue, 5 Dec 2023 17:18:22 +1100
Subject: [PATCH 13/15] review comment, hstack

---
 sklearn/multioutput.py | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 528751011654a..854c4343cca8b 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -664,12 +664,10 @@ def _get_predictions(self, X, *, output_method):
         # `RegressorChain` does not have a `chain_method_` parameter so we
         # default to "predict"
         chain_method = getattr(self, "chain_method_", "predict")
+        hstack = sp.hstack if sp.issparse(X) else np.hstack
         for chain_idx, estimator in enumerate(self.estimators_):
             previous_predictions = Y_feature_chain[:, :chain_idx]
-            if sp.issparse(X):
-                X_aug = sp.hstack((X, previous_predictions))
-            else:
-                X_aug = np.hstack((X, previous_predictions))
+            X_aug = hstack((X, previous_predictions))
 
             feature_predictions, _ = _get_response_values(
                 estimator, X_aug, response_method=chain_method,
@@ -781,6 +779,7 @@ def fit(self, X, Y, **fit_params):
                     cv=self.cv,
                     method=chain_method,
                 )
+                # `predict_proba` output is 2D, we use only output for classes[-1]
                 if cv_result.ndim > 1:
                     cv_result = cv_result[:, 1]
                 if sp.issparse(X_aug):

From 443bcdbd76d6449a8d33832b409177460c8c3d13 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Tue, 5 Dec 2023 20:38:10 +1100
Subject: [PATCH 14/15] black

---
 sklearn/multioutput.py            | 11 ++++++++---
 sklearn/tests/test_multioutput.py |  3 ++-
 2 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index 854c4343cca8b..3ee89d40058c1 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -670,12 +670,16 @@ def _get_predictions(self, X, *, output_method):
             X_aug = hstack((X, previous_predictions))
 
             feature_predictions, _ = _get_response_values(
-                estimator, X_aug, response_method=chain_method,
+                estimator,
+                X_aug,
+                response_method=chain_method,
             )
             Y_feature_chain[:, chain_idx] = feature_predictions
 
             output_predictions, _ = _get_response_values(
-                estimator, X_aug, response_method=output_method,
+                estimator,
+                X_aug,
+                response_method=output_method,
             )
             Y_output_chain[:, chain_idx] = output_predictions
 
@@ -749,7 +753,8 @@ def fit(self, X, Y, **fit_params):
 
         if hasattr(self, "chain_method"):
             chain_method = _check_response_method(
-                self.base_estimator, self.chain_method,
+                self.base_estimator,
+                self.chain_method,
             ).__name__
             self.chain_method_ = chain_method
         else:
diff --git a/sklearn/tests/test_multioutput.py b/sklearn/tests/test_multioutput.py
index d16d29d460776..7ad1c797af257 100644
--- a/sklearn/tests/test_multioutput.py
+++ b/sklearn/tests/test_multioutput.py
@@ -497,7 +497,8 @@ def test_classifier_chain_fit_and_predict_with_linear_svc(chain_method):
     # Fit classifier chain and verify predict performance using LinearSVC
     X, Y = generate_multilabel_dataset_with_correlations()
     classifier_chain = ClassifierChain(
-        LinearSVC(dual="auto"), chain_method=chain_method,
+        LinearSVC(dual="auto"),
+        chain_method=chain_method,
     ).fit(X, Y)
 
     Y_pred = classifier_chain.predict(X)

From 969d8892a6e78ae91c028374f3479d0c94a9cb86 Mon Sep 17 00:00:00 2001
From: Lucy Liu <jliu176@gmail.com>
Date: Fri, 23 Feb 2024 11:49:36 +1100
Subject: [PATCH 15/15] update v

---
 doc/whats_new/v1.4.rst | 6 ------
 doc/whats_new/v1.5.rst | 6 ++++++
 sklearn/multioutput.py | 2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/doc/whats_new/v1.4.rst b/doc/whats_new/v1.4.rst
index 76737a1722e4a..7092c53da1a27 100644
--- a/doc/whats_new/v1.4.rst
+++ b/doc/whats_new/v1.4.rst
@@ -865,12 +865,6 @@ Changelog
 - |Enhancement| Add method `predict_log_proba` to :class:`multioutput.ClassifierChain`.
   :pr:`27720` by :user:`Guillaume Lemaitre <glemaitre>`.
 
-:mod:`sklearn.multioutput`
-..........................
-
-- |Enhancement| `chain_method` parameter added to `:class:``multioutput.ClassifierChain`.
-  :pr:`27700` by :user:`Lucy Liu <lucyleeow>`.
-
 :mod:`sklearn.neighbors`
 ........................
 
diff --git a/doc/whats_new/v1.5.rst b/doc/whats_new/v1.5.rst
index 53f0fbd8a74e8..d70c9cc2f1f23 100644
--- a/doc/whats_new/v1.5.rst
+++ b/doc/whats_new/v1.5.rst
@@ -158,6 +158,12 @@ Changelog
 - |Enhancement| :term:`CV splitters <CV splitter>` that ignores the group parameter now
   raises a warning when groups are passed in to :term:`split`. :pr:`28210` by
 
+:mod:`sklearn.multioutput`
+..........................
+
+- |Enhancement| `chain_method` parameter added to `:class:``multioutput.ClassifierChain`.
+  :pr:`27700` by :user:`Lucy Liu <lucyleeow>`.
+
 :mod:`sklearn.pipeline`
 .......................
 
diff --git a/sklearn/multioutput.py b/sklearn/multioutput.py
index f593497336da9..64649007d6f24 100644
--- a/sklearn/multioutput.py
+++ b/sklearn/multioutput.py
@@ -871,7 +871,7 @@ class ClassifierChain(MetaEstimatorMixin, ClassifierMixin, _BaseChain):
           preference. The method used corresponds to the first method in
           the list that is implemented by `base_estimator`.
 
-        .. versionadded:: 1.4
+        .. versionadded:: 1.5
 
     random_state : int, RandomState instance or None, optional (default=None)
         If ``order='random'``, determines random number generation for the