MNT Change warning type and validation location for contamination in LOF (scikit-learn#11634)

Author: glemaitre

sklearn/neighbors/lof.py

@@ -100,6 +100,10 @@ class LocalOutlierFactor(NeighborsBase, KNeighborsMixin, UnsupervisedMixin,
  threshold on the decision function. If "auto", the decision function
  threshold is determined as in the original paper.
 
+ .. versionchanged:: 0.20
+ The default value of ``contamination`` will change from 0.1 in 0.20
+ to ``'auto'`` in 0.22.
+
  novelty : boolean, default False
  By default, LocalOutlierFactor is only meant to be used for outlier
  detection (novelty=False). Set novelty to True if you want to use
@@ -150,12 +154,6 @@ def __init__(self, n_neighbors=20, algorithm='auto', leaf_size=30,
  algorithm=algorithm,
  leaf_size=leaf_size, metric=metric, p=p,
  metric_params=metric_params, n_jobs=n_jobs)
-
- if contamination == "legacy":
- warnings.warn('default contamination parameter 0.1 will change '
- 'in version 0.22 to "auto". This will change the '
- 'predict method behavior.',
- DeprecationWarning)
  self.contamination = contamination
  self.novelty = novelty
 
@@ -224,10 +222,19 @@ def fit(self, X, y=None):
  -------
  self : object
  """
- if self.contamination not in ["auto", "legacy"]: # rm legacy in 0.22
- if not(0. < self.contamination <= .5):
+ if self.contamination == "legacy":
+ warnings.warn('default contamination parameter 0.1 will change '
+ 'in version 0.22 to "auto". This will change the '
+ 'predict method behavior.',
+ FutureWarning)
+ self._contamination = 0.1
+ else:
+ self._contamination = self.contamination
+
+ if self._contamination != 'auto':
+ if not(0. < self._contamination <= .5):
  raise ValueError("contamination must be in (0, 0.5], "
- "got: %f" % self.contamination)
+ "got: %f" % self._contamination)
 
  super(LocalOutlierFactor, self).fit(X)
 
@@ -251,15 +258,12 @@ def fit(self, X, y=None):
 
  self.negative_outlier_factor_ = -np.mean(lrd_ratios_array, axis=1)
 
- if self.contamination == "auto":
+ if self._contamination == "auto":
  # inliers score around -1 (the higher, the less abnormal).
  self.offset_ = -1.5
- elif self.contamination == "legacy": # to rm in 0.22
- self.offset_ = scoreatpercentile(
- self.negative_outlier_factor_, 100. * 0.1)
  else:
  self.offset_ = scoreatpercentile(
- self.negative_outlier_factor_, 100. * self.contamination)
+ self.negative_outlier_factor_, 100. * self._contamination)
 
  return self
 

sklearn/neighbors/tests/test_lof.py

@@ -3,6 +3,8 @@
 # License: BSD 3 clause
 
 from math import sqrt
+
+import pytest
 import numpy as np
 from sklearn import neighbors
 
@@ -32,8 +34,9 @@
 iris.target = iris.target[perm]
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_lof():
  # Toy sample (the last two samples are outliers):
  X = [[-2, -1], [-1, -1], [-1, -2], [1, 1], [1, 2], [2, 1], [5, 3], [-4, 2]]
@@ -53,8 +56,9 @@ def test_lof():
  assert_array_equal(clf.fit_predict(X), 6 * [1] + 2 * [-1])
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_lof_performance():
  # Generate train/test data
  rng = check_random_state(2)
@@ -76,8 +80,9 @@ def test_lof_performance():
  assert_greater(roc_auc_score(y_test, y_pred), .99)
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_lof_values():
  # toy samples:
  X_train = [[1, 1], [1, 2], [2, 1]]
@@ -99,8 +104,9 @@ def test_lof_values():
  assert_array_almost_equal(-clf2.score_samples([[1., 1.]]), [s_1])
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_lof_precomputed(random_state=42):
  """Tests LOF with a distance matrix."""
  # Note: smaller samples may result in spurious test success
@@ -126,8 +132,9 @@ def test_lof_precomputed(random_state=42):
  assert_array_almost_equal(pred_X_Y, pred_D_Y)
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_n_neighbors_attribute():
  X = iris.data
  clf = neighbors.LocalOutlierFactor(n_neighbors=500).fit(X)
@@ -140,8 +147,9 @@ def test_n_neighbors_attribute():
  assert_equal(clf.n_neighbors_, X.shape[0] - 1)
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_score_samples():
  X_train = [[1, 1], [1, 2], [2, 1]]
  clf1 = neighbors.LocalOutlierFactor(n_neighbors=2,
@@ -163,8 +171,9 @@ def test_contamination():
  assert_raises(ValueError, clf.fit, X)
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_novelty_errors():
  X = iris.data
 
@@ -182,8 +191,9 @@ def test_novelty_errors():
  assert_raises_regex(AttributeError, msg, getattr, clf, 'fit_predict')
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_novelty_training_scores():
  # check that the scores of the training samples are still accessible
  # when novelty=True through the negative_outlier_factor_ attribute
@@ -202,8 +212,9 @@ def test_novelty_training_scores():
  assert_array_almost_equal(scores_1, scores_2)
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_hasattr_prediction():
  # check availability of prediction methods depending on novelty value.
  X = [[1, 1], [1, 2], [2, 1]]
@@ -225,16 +236,19 @@ def test_hasattr_prediction():
  assert not hasattr(clf, 'score_samples')
 
 
-@ignore_warnings(category=DeprecationWarning)
-# contamination changed to 'auto' 0.22
+@pytest.mark.filterwarnings(
+ 'ignore:default contamination parameter 0.1:FutureWarning')
+# XXX: Remove in 0.22
 def test_novelty_true_common_tests():
+
  # the common tests are run for the default LOF (novelty=False).
  # here we run these common tests for LOF when novelty=True
  check_estimator(neighbors.LocalOutlierFactor(novelty=True))
 
 
-def test_deprecation():
- assert_warns_message(DeprecationWarning,
+def test_contamination_future_warning():
+ X = [[1, 1], [1, 2], [2, 1]]
+ assert_warns_message(FutureWarning,
  'default contamination parameter 0.1 will change '
  'in version 0.22 to "auto"',
- neighbors.LocalOutlierFactor, )
+ neighbors.LocalOutlierFactor().fit, X)