def transform_problem(multilabel_dataset, feature_selector, classifier) -> []:
partitioned_datasets = partition_dataset(multilabel_dataset)
fitted_feature_selectors = \
Stream(partitioned_datasets) \
.map(lambda elem: select_features(elem, feature_selector)).as_list()
partitioned_results = \
Stream(fitted_feature_selectors) \
.map(lambda elem: classify(elem, classifier)) \
.as_list()
return combine_partitioned_results(multilabel_dataset, partitioned_results)
def partition_dataset(multilabel_dataset):
partitioned_labels = partition_labels(multilabel_dataset)
return \
Stream(partitioned_labels) \
.map(lambda label_partition: augment_label(multilabel_dataset, label_partition)) \
.as_list()
def test_chaining(self):
expected = [3, 5]
actual = \
Stream([1, 2, 3, 4]) \
.filter(lambda elem: elem%2 == 0) \
.map(lambda elem: elem+1) \
.as_list() \
assert expected == actual, "should be [3, 5]"
def test_append_elem(self):
expected = [1, 2, 3, 4]
actual = \
Stream([1, 2]) \
.append_elem(3) \
.append_elem(4) \
.as_list()
print("actual: ", actual)
assert expected == actual, "should be [1, 2, 3, 4]"
def test_append_list(self):
expected = [1, 2, 3, 4, 5, 6]
actual = \
Stream([1, 2]) \
.append_list([3, 4]) \
.append_list([5, 6]) \
.as_list()
print("actual: ", actual)
assert expected == actual, "should be [1, 2, 3, 4, 5, 6]"
def test_map(self):
expected = [2, 3, 4, 5]
actual = Stream([1, 2, 3, 4]).map(lambda elem: elem + 1).as_list()
assert expected == actual, "should be [2, 3, 4, 5]"
def test_filter(self):
expected = [2, 4]
actual = Stream([1, 2, 3,
4]).filter(lambda elem: elem % 2 == 0).as_list()
assert expected == actual, "should be all even numbers"
def create_labels_dataframe(multilabel_dataset, partitioned_results):
labels = get_total_labels_test(multilabel_dataset)
label_names = Stream(labels).map(lambda elem: elem[0]).as_list()
return DataFrame(partitioned_results, index=label_names).transpose()
def create_features_dataframe(multilabel_dataset):
features = get_instance_features_test(multilabel_dataset)
total_features = get_total_features_test(multilabel_dataset)
feature_names = Stream(total_features).map(lambda elem: elem[0]).as_list()
return DataFrame(features, columns=feature_names)