Merge remote-tracking branch 'upstream/pr/6182'

Author: MechCoder

doc/whats_new.rst

@@ -102,6 +102,9 @@ Enhancements
 Bug fixes
 .........
 
+    - :class:`StratifiedKFold` now raises error if all n_labels for individual classes is less than n_folds.
+      (`#6182 <https://github.com/scikit-learn/scikit-learn/pull/6182>`_) by `Devashish Deshpande`_.
+
     - :class:`RandomizedPCA` default number of `iterated_power` is 2 instead of 3.
       This is a speed up with a minor precision decrease. (`#5141 <https://github.com/scikit-learn/scikit-learn/pull/5141>`_) by `Giorgio Patrini`_.
 

sklearn/cross_validation.py

@@ -519,6 +519,10 @@ def __init__(self, y, n_folds=3, shuffle=False,
         unique_labels, y_inversed = np.unique(y, return_inverse=True)
         label_counts = bincount(y_inversed)
         min_labels = np.min(label_counts)
+        if np.all(self.n_folds > label_counts):
+            raise ValueError("All the n_labels for individual classes"
+                             " are less than %d folds."
+                             % (self.n_folds))
         if self.n_folds > min_labels:
             warnings.warn(("The least populated class in y has only %d"
                            " members, which is too few. The minimum"

sklearn/model_selection/_split.py

@@ -564,6 +564,10 @@ def _make_test_folds(self, X, y=None, labels=None):
         unique_y, y_inversed = np.unique(y, return_inverse=True)
         y_counts = bincount(y_inversed)
         min_labels = np.min(y_counts)
+        if np.all(self.n_folds > y_counts):
+            raise ValueError("All the n_labels for individual classes"
+                             " are less than %d folds."
+                             % (self.n_folds))
         if self.n_folds > min_labels:
             warnings.warn(("The least populated class in y has only %d"
                            " members, which is too few. The minimum"

sklearn/model_selection/tests/test_split.py

@@ -21,6 +21,7 @@
 from sklearn.utils.testing import assert_array_almost_equal
 from sklearn.utils.testing import assert_array_equal
 from sklearn.utils.testing import assert_warns_message
+from sklearn.utils.testing import assert_raise_message
 from sklearn.utils.testing import ignore_warnings
 from sklearn.utils.validation import _num_samples
 from sklearn.utils.mocking import MockDataFrame
@@ -206,7 +207,7 @@ def test_kfold_valueerrors():
 
     # Check that a warning is raised if the least populated class has too few
     # members.
-    y = np.array([3, 3, -1, -1, 2])
+    y = np.array([3, 3, -1, -1, 3])
 
     skf_3 = StratifiedKFold(3)
     assert_warns_message(Warning, "The least populated class",
@@ -219,11 +220,21 @@ def test_kfold_valueerrors():
         warnings.simplefilter("ignore")
         check_cv_coverage(skf_3, X2, y, labels=None, expected_n_iter=3)
 
+    # Check that errors are raised if all n_labels for individual
+    # classes are less than n_folds.
+    y = np.array([3, 3, -1, -1, 2])
+
+    assert_raises(ValueError, next, skf_3.split(X2, y))
+
     # Error when number of folds is <= 1
     assert_raises(ValueError, KFold, 0)
     assert_raises(ValueError, KFold, 1)
-    assert_raises(ValueError, StratifiedKFold, 0)
-    assert_raises(ValueError, StratifiedKFold, 1)
+    error_string = ("k-fold cross-validation requires at least one"
+                    " train/test split")
+    assert_raise_message(ValueError, error_string,
+                         StratifiedKFold, 0)
+    assert_raise_message(ValueError, error_string,
+                         StratifiedKFold, 1)
 
     # When n_folds is not integer:
     assert_raises(ValueError, KFold, 1.5)

sklearn/tests/test_cross_validation.py

@@ -19,6 +19,7 @@
 from sklearn.utils.testing import assert_array_almost_equal
 from sklearn.utils.testing import assert_array_equal
 from sklearn.utils.testing import assert_warns_message
+from sklearn.utils.testing import assert_raise_message
 from sklearn.utils.testing import ignore_warnings
 from sklearn.utils.mocking import CheckingClassifier, MockDataFrame
 
@@ -160,7 +161,7 @@ def test_kfold_valueerrors():
 
     # Check that a warning is raised if the least populated class has too few
     # members.
-    y = [3, 3, -1, -1, 2]
+    y = [3, 3, -1, -1, 3]
 
     cv = assert_warns_message(Warning, "The least populated class",
                               cval.StratifiedKFold, y, 3)
@@ -170,11 +171,21 @@ def test_kfold_valueerrors():
     # side of the split at each split
     check_cv_coverage(cv, expected_n_iter=3, n_samples=len(y))
 
+    # Check that errors are raised if all n_labels for individual
+    # classes are less than n_folds.
+    y = [3, 3, -1, -1, 2]
+
+    assert_raises(ValueError, cval.StratifiedKFold, y, 3)
+
     # Error when number of folds is <= 1
     assert_raises(ValueError, cval.KFold, 2, 0)
     assert_raises(ValueError, cval.KFold, 2, 1)
-    assert_raises(ValueError, cval.StratifiedKFold, y, 0)
-    assert_raises(ValueError, cval.StratifiedKFold, y, 1)
+    error_string = ("k-fold cross validation requires at least one"
+                    " train / test split")
+    assert_raise_message(ValueError, error_string,
+                         cval.StratifiedKFold, y, 0)
+    assert_raise_message(ValueError, error_string,
+                         cval.StratifiedKFold, y, 1)
 
     # When n is not integer:
     assert_raises(ValueError, cval.KFold, 2.5, 2)