ENH: multi-output support for BaggingRegressor

dohmatob · dohmatob · commit e3acfa413fcd · 2017-06-09T06:51:56.000+02:00
diff --git a/doc/whats_new.rst b/doc/whats_new.rst
@@ -823,6 +823,9 @@ Trees and ensembles
      :class:`ensemble.VotingClassifier` to fit underlying estimators in parallel.
      :issue:`5805` by :user:`Ibraim Ganiev <olologin>`.
 
+   - :class:`ensemble.BaggingRegressor` now supports multi-output targets.
+     By :user:`Elvis Dohmatob <dohmatob>`.
+	    
 Linear, kernelized and related models
 
    - In :class:`linear_model.LogisticRegression`, the SAG solver is now
diff --git a/sklearn/ensemble/bagging.py b/sklearn/ensemble/bagging.py
@@ -280,7 +280,7 @@ def _fit(self, X, y, max_samples=None, max_depth=None, sample_weight=None):
         random_state = check_random_state(self.random_state)
 
         # Convert data
-        X, y = check_X_y(X, y, ['csr', 'csc'])
+        X, y = check_X_y(X, y, ['csr', 'csc'], multi_output=True)
         if sample_weight is not None:
             sample_weight = check_array(sample_weight, ensure_2d=False)
             check_consistent_length(y, sample_weight)
@@ -390,8 +390,9 @@ def _set_oob_score(self, X, y):
         """Calculate out of bag predictions and score."""
 
     def _validate_y(self, y):
-        # Default implementation
-        return column_or_1d(y, warn=True)
+        # Default implementation. We skip column_or_1d and similar checks
+        # in order to make the code support multi-output targets.
+        return y
 
     def _get_estimators_indices(self):
         # Get drawn indices along both sample and feature axes
diff --git a/sklearn/ensemble/tests/test_bagging.py b/sklearn/ensemble/tests/test_bagging.py
@@ -7,7 +7,7 @@
 
 import numpy as np
 
-from sklearn.base import BaseEstimator
+from sklearn.base import BaseEstimator, clone
 
 from sklearn.utils.testing import assert_array_equal
 from sklearn.utils.testing import assert_array_almost_equal
@@ -23,7 +23,7 @@
 from sklearn.dummy import DummyClassifier, DummyRegressor
 from sklearn.model_selection import GridSearchCV, ParameterGrid
 from sklearn.ensemble import BaggingClassifier, BaggingRegressor
-from sklearn.linear_model import Perceptron, LogisticRegression
+from sklearn.linear_model import Perceptron, LogisticRegression, Ridge
 from sklearn.neighbors import KNeighborsClassifier, KNeighborsRegressor
 from sklearn.tree import DecisionTreeClassifier, DecisionTreeRegressor
 from sklearn.svm import SVC, SVR
@@ -740,3 +740,24 @@ def test_set_oob_score_label_encoding():
     x3 = BaggingClassifier(oob_score=True,
                            random_state=random_state).fit(X, Y3).oob_score_
     assert_equal([x1, x2], [x3, x3])
+
+
+def test_multi_output_regressor():
+    # Check singleton ensembles.
+    rng = check_random_state(0)
+    X_train, X_test, y_train, y_test = train_test_split(boston.data,
+                                                        boston.target,
+                                                        random_state=rng)
+
+    reg1 = BaggingRegressor(base_estimator=Ridge(), n_estimators=10,
+                            bootstrap=False, bootstrap_features=False,
+                            random_state=rng)
+    assert_false(reg1.multi_output)
+    reg2 = clone(reg1)
+    reg1.fit(X_train, y_train)
+    for n_targets in [1, 2]:
+        y_train_ = np.ndarray((len(y_train), n_targets))
+        y_train_.T[:] = y_train.copy()
+        reg2.fit(X_train, y_train_)
+        assert_array_almost_equal(reg1.predict(X_test),
+                                  reg2.predict(X_test)[:, 0], decimal=10)
diff --git a/sklearn/utils/estimator_checks.py b/sklearn/utils/estimator_checks.py
@@ -64,7 +64,8 @@
                 'MultiTaskElasticNetCV', 'MultiTaskLasso', 'MultiTaskLassoCV',
                 'OrthogonalMatchingPursuit', 'PLSCanonical', 'PLSRegression',
                 'RANSACRegressor', 'RadiusNeighborsRegressor',
-                'RandomForestRegressor', 'Ridge', 'RidgeCV']
+                'RandomForestRegressor', 'Ridge', 'RidgeCV',
+                "BaggingRegressor"]
 
 
 def _yield_non_meta_checks(name, Estimator):
