FIX: raise an error when multilabel does not encode multiclass (#490)

Author: glemaitre

doc/whats_new.rst

@@ -4,10 +4,10 @@
 Release history
 ===============
 
-.. include:: whats_new/v0.0.4.rst
+.. include:: whats_new/v0.4.rst
 
-.. include:: whats_new/v0.0.3.rst
+.. include:: whats_new/v0.3.rst
 
-.. include:: whats_new/v0.0.2.rst
+.. include:: whats_new/v0.2.rst
 
-.. include:: whats_new/v0.0.1.rst
+.. include:: whats_new/v0.1.rst

doc/whats_new/v0.0.1.rst renamed to doc/whats_new/v0.1.rst

@@ -14,6 +14,10 @@ Bug fixes
   deviation.
   By :user:`Guillaume Lemaitre <glemaitre>` in :issue:`491`.
 
+- Raise an error when passing target  which is not supported, i.e. regression
+  target or multilabel targets. Imbalanced-learn does not support this case.
+  By :user:`Guillaume Lemaitre <glemaitre>` in :issue:`490`.
+
 Version 0.4
 ===========
 

doc/whats_new/v0.0.2.rst renamed to doc/whats_new/v0.2.rst

@@ -15,6 +15,7 @@
 from sklearn.externals import six
 from sklearn.preprocessing import label_binarize
 from sklearn.utils import check_X_y
+from sklearn.utils.multiclass import check_classification_targets
 
 from .utils import check_sampling_strategy, check_target_type
 from .utils.deprecation import deprecate_parameter
@@ -77,6 +78,7 @@ def fit_resample(self, X, y):
         """
         self._deprecate_ratio()
 
+        check_classification_targets(y)
         X, y, binarize_y = self._check_X_y(X, y)
 
         self.sampling_strategy_ = check_sampling_strategy(

doc/whats_new/v0.0.3.rst renamed to doc/whats_new/v0.3.rst

@@ -127,9 +127,12 @@ def test_smotenc_check_target_type():
     y = np.linspace(0, 1, 30)
     smote = SMOTENC(categorical_features=categorical_features,
                     random_state=0)
-    with pytest.warns(UserWarning, match='should be of types'):
+    with pytest.raises(ValueError, match="Unknown label type: 'continuous'"):
+        smote.fit_resample(X, y)
+    rng = np.random.RandomState(42)
+    y = rng.randint(2, size=(20, 3))
+    with pytest.raises(ValueError, match="'y' should encode the multiclass"):
         smote.fit_resample(X, y)
-
 
 def test_smotenc_samplers_one_label():
     X, _, categorical_features = data_heterogneous_unordered()

doc/whats_new/v0.0.4.rst renamed to doc/whats_new/v0.4.rst

@@ -86,17 +86,14 @@ def check_target_type(y, indicate_one_vs_all=False):
 
     """
     type_y = type_of_target(y)
-    if type_y not in TARGET_KIND:
-        # FIXME: perfectly we should raise an error but the sklearn API does
-        # not allow for it
-        warnings.warn("'y' should be of types {} only. Got {} instead.".format(
-            TARGET_KIND, type_of_target(y)))
-
-    if indicate_one_vs_all:
-        return (y.argmax(axis=1) if type_y == 'multilabel-indicator' else y,
-                type_y == 'multilabel-indicator')
-    else:
-        return y.argmax(axis=1) if type_y == 'multilabel-indicator' else y
+    if type_y == 'multilabel-indicator':
+        if np.any(y.sum(axis=1) > 1):
+            raise ValueError(
+                "When 'y' corresponds to '{}', 'y' should encode the "
+                "multiclass (a single 1 by row).".format(type_y))
+        y = y.argmax(axis=1)
+
+    return (y, type_y == 'multilabel-indicator') if indicate_one_vs_all else y
 
 
 def _sampling_strategy_all(y, sampling_type):

imblearn/base.py

@@ -33,8 +33,6 @@
 from imblearn.over_sampling import SMOTE
 from imblearn.under_sampling import NearMiss, ClusterCentroids
 
-from imblearn.utils.testing import warns
-
 DONT_SUPPORT_RATIO = ['SVMSMOTE', 'BorderlineSMOTE']
 SUPPORT_STRING = ['RandomUnderSampler', 'RandomOverSampler']
 HAVE_SAMPLE_INDICES = [
@@ -54,7 +52,6 @@ def monkey_patch_check_dtype_object(name, estimator_orig):
     X = rng.rand(40, 10).astype(object)
     y = np.array([0] * 10 + [1] * 30, dtype=np.int)
     estimator = clone(estimator_orig)
-
     estimator.fit(X, y)
 
     try:
@@ -123,14 +120,20 @@ def check_estimator(Estimator, run_sampler_tests=True):
 
 
 def check_target_type(name, Estimator):
+    # should raise warning if the target is continuous (we cannot raise error)
     X = np.random.random((20, 2))
     y = np.linspace(0, 1, 20)
     estimator = Estimator()
     # FIXME: in 0.6 set the random_state for all
     if name not in DONT_HAVE_RANDOM_STATE:
         set_random_state(estimator)
-    with warns(UserWarning, match='should be of types'):
-        estimator.fit(X, y)
+    with pytest.raises(ValueError, match="Unknown label type: 'continuous'"):
+        estimator.fit_resample(X, y)
+    # if the target is multilabel then we should raise an error
+    rng = np.random.RandomState(42)
+    y = rng.randint(2, size=(20, 3))
+    with pytest.raises(ValueError, match="'y' should encode the multiclass"):
+        estimator.fit_resample(X, y)
 
 
 def check_samplers_one_label(name, Sampler):
@@ -139,7 +142,7 @@ def check_samplers_one_label(name, Sampler):
     X = np.random.random((20, 2))
     y = np.zeros(20)
     try:
-        sampler.fit(X, y)
+        sampler.fit_resample(X, y)
     except ValueError as e:
         if 'class' not in repr(e):
             print(error_string_fit, Sampler, e)
@@ -157,15 +160,15 @@ def check_samplers_fit(name, Sampler):
     sampler = Sampler()
     X = np.random.random((30, 2))
     y = np.array([1] * 20 + [0] * 10)
-    sampler.fit(X, y)
+    sampler.fit_resample(X, y)
     assert hasattr(sampler, 'sampling_strategy_'), \
         "No fitted attribute sampling_strategy_"
 
 
 def check_samplers_fit_resample(name, Sampler):
     sampler = Sampler()
     X, y = make_classification(n_samples=1000, n_classes=3, n_informative=4,
-                                weights=[0.2, 0.3, 0.5], random_state=0)
+                               weights=[0.2, 0.3, 0.5], random_state=0)
     target_stats = Counter(y)
     X_res, y_res = sampler.fit_resample(X, y)
     if isinstance(sampler, BaseOverSampler):

imblearn/over_sampling/tests/test_smote_nc.py

@@ -3,6 +3,7 @@
 
 from sklearn.base import BaseEstimator
 from sklearn.utils import check_X_y
+from sklearn.utils.multiclass import check_classification_targets
 
 from imblearn.base import BaseSampler
 from imblearn.utils.estimator_checks import check_estimator
@@ -17,6 +18,8 @@ def fit(self, X, y):
         return self
 
     def fit_resample(self, X, y):
+        check_classification_targets(y)
+        self.fit(X, y)
         return X, y
 
 
@@ -27,36 +30,15 @@ def fit(self, X, y):
         X, y = check_X_y(X, y, accept_sparse=True)
         return self
 
-    def fit_resample(self, X, y):
-        self.fit(X, y)
-        return X, y
-
 
 class NoAcceptingSparseSampler(BaseBadSampler):
     """Sampler which does not accept sparse matrix."""
     def fit(self, X, y):
-        X, y = check_X_y(X, y, accept_sparse=False)
-        y, _ = check_target_type(y, indicate_one_vs_all=True)
-        self.sampling_strategy_ = 'sampling_strategy_'
-        return self
-
-    def fit_resample(self, X, y):
-        self.fit(X, y)
-        return X, y
-
-
-class NotTransformingTargetOvR(BaseBadSampler):
-    """Sampler which does not transform OvR enconding."""
-    def fit(self, X, y):
-        X, y = check_X_y(X, y, accept_sparse=True)
         y, _ = check_target_type(y, indicate_one_vs_all=True)
+        X, y = check_X_y(X, y, accept_sparse=False)
         self.sampling_strategy_ = 'sampling_strategy_'
         return self
 
-    def fit_resample(self, X, y):
-        self.fit(X, y)
-        return X, y
-
 
 class NotPreservingDtypeSampler(BaseSampler):
     _sampling_type = 'bypass'
@@ -72,7 +54,6 @@ def _fit_resample(self, X, y):
     [(BaseBadSampler, AssertionError, "TypeError not raised by fit"),
      (NotFittedSampler, AssertionError, "No fitted attribute"),
      (NoAcceptingSparseSampler, TypeError, "A sparse matrix was passed"),
-     (NotTransformingTargetOvR, ValueError, "bad input shape"),
      (NotPreservingDtypeSampler, AssertionError, "X dytype is not preserved")]
 )
 def test_check_estimator(Estimator, err_type, err_msg):