def test_if_dense_classification_works_on_non_dense_base_classifier(
self):
classifier = BinaryRelevance(classifier=Keras(
create_model_single_class, False, KERAS_PARAMS),
require_dense=[True, True])
self.assertClassifierWorksWithSparsity(classifier, 'dense')
def test_model_selection_works(self):
for x, y in self.get_multilabel_data_for_tests('dense'):
parameters = {
'classifier': [LabelPowerset(),
BinaryRelevance()],
'clusterer': [RandomLabelSpaceClusterer(None, None, False)],
'clusterer__cluster_size': list(range(2, 3)),
'clusterer__cluster_count': [3],
'clusterer__allow_overlap': [False],
'classifier__classifier': [MultinomialNB()],
'classifier__classifier__alpha': [0.7, 1.0],
}
clf = GridSearchCV(LabelSpacePartitioningClassifier(),
parameters,
scoring='f1_macro')
clf.fit(x, y)
for p in list(parameters.keys()):
self.assertIn(p, clf.best_params_)
self.assertIsNotNone(clf.best_score_)
def test_if_works_with_cross_validation(self):
classifier = BinaryRelevance(
classifier=GaussianNB(), require_dense=[True, True])
self.assertClassifierWorksWithCV(classifier)
def test_if_dense_classification_works_on_dense_base_classifier(self):
classifier = BinaryRelevance(
classifier=GaussianNB(), require_dense=[True, True])
self.assertClassifierWorksWithSparsity(classifier, 'dense')
self.assertClassifierPredictsProbabilities(classifier, 'dense')
def test_if_sparse_classification_works_on_non_dense_base_classifier(self):
classifier = BinaryRelevance(
classifier=SVC(probability=True), require_dense=[False, True])
self.assertClassifierWorksWithSparsity(classifier, 'sparse')
self.assertClassifierPredictsProbabilities(classifier, 'sparse')