Test multilabel classifier on random dataset.

sklearn/tests/test_multiclass.py

@@ -5,6 +5,7 @@
 from numpy.testing import assert_array_almost_equal
 from numpy.testing import assert_array_equal
 from nose.tools import assert_equal
+from nose.tools import assert_almost_equal
 from nose.tools import assert_true
 from nose.tools import assert_raises
 
@@ -23,6 +24,30 @@
 n_classes = 3
 
 
+# FIXME: - should use sets
+#        - should move to metrics module
+def multilabel_precision(Y_true, Y_pred):
+    n_predictions = 0
+    n_correct = 0
+    for i in range(len(Y_true)):
+        n_predictions += len(Y_pred[i])
+        for label in Y_pred[i]:
+            if label in Y_true[i]:
+                n_correct += 1
+    return float(n_correct) / n_predictions
+
+
+def multilabel_recall(Y_true, Y_pred):
+    n_labels = 0
+    n_correct = 0
+    for i in range(len(Y_true)):
+        n_labels += len(Y_true[i])
+        for label in Y_pred[i]:
+            if label in Y_true[i]:
+                n_correct += 1
+    return float(n_correct) / n_labels
+
+
 def test_ovr_exceptions():
     ovr = OneVsRestClassifier(LinearSVC())
     assert_raises(ValueError, ovr.predict, [])
@@ -64,6 +89,22 @@ def test_ovr_multilabel():
         assert_array_equal(y_pred, [0, 1, 1])
 
 
+def test_ovr_multilabel_dataset():
+    base_clf = MultinomialNB()
+    X, Y = datasets.make_multilabel_classification(n_samples=100,
+                                                   n_features=20,
+                                                   n_classes=5,
+                                                   n_labels=2,
+                                                   length=50,
+                                                   random_state=0)
+    X_train, Y_train = X[:80], Y[:80]
+    X_test, Y_test = X[80:], Y[80:]
+    clf = OneVsRestClassifier(base_clf).fit(X_train, Y_train)
+    Y_pred = clf.predict(X_test)
+    assert_almost_equal(multilabel_precision(Y_test, Y_pred), 0.80, places=2)
+    assert_almost_equal(multilabel_recall(Y_test, Y_pred), 0.80, places=2)
+
+
 def test_ovr_gridsearch():
     ovr = OneVsRestClassifier(LinearSVC())
     Cs = [0.1, 0.5, 0.8]