def test_multilabel_binarizer_given_classes():
inp = [(2, 3), (1,), (1, 2)]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 0, 1]])
# fit_transform()
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, [1, 3, 2])
# fit().transform()
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, [1, 3, 2])
# ensure works with extra class
mlb = MultiLabelBinarizer(classes=[4, 1, 3, 2])
assert_array_equal(mlb.fit_transform(inp),
np.hstack(([[0], [0], [0]], indicator_mat)))
assert_array_equal(mlb.classes_, [4, 1, 3, 2])
# ensure fit is no-op as iterable is not consumed
inp = iter(inp)
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
# ensure a ValueError is thrown if given duplicate classes
err_msg = "The classes argument contains duplicate classes. Remove " \
"these duplicates before passing them to MultiLabelBinarizer."
mlb = MultiLabelBinarizer(classes=[1, 3, 2, 3])
assert_raise_message(ValueError, err_msg, mlb.fit, inp)
def test_multilabel_binarizer_given_classes():
inp = [(2, 3), (1,), (1, 2)]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 0, 1]])
# fit_transform()
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, [1, 3, 2])
# fit().transform()
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, [1, 3, 2])
# ensure works with extra class
mlb = MultiLabelBinarizer(classes=[4, 1, 3, 2])
assert_array_equal(mlb.fit_transform(inp),
np.hstack(([[0], [0], [0]], indicator_mat)))
assert_array_equal(mlb.classes_, [4, 1, 3, 2])
# ensure fit is no-op as iterable is not consumed
inp = iter(inp)
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
# ensure a ValueError is thrown if given duplicate classes
err_msg = "The classes argument contains duplicate classes. Remove " \
"these duplicates before passing them to MultiLabelBinarizer."
mlb = MultiLabelBinarizer(classes=[1, 3, 2, 3])
assert_raise_message(ValueError, err_msg, mlb.fit, inp)
def test_multilabel_binarizer_given_classes():
inp = [(2, 3), (1,), (1, 2)]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 0, 1]])
# fit_transform()
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, [1, 3, 2])
# fit().transform()
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, [1, 3, 2])
# ensure works with extra class
mlb = MultiLabelBinarizer(classes=[4, 1, 3, 2])
assert_array_equal(mlb.fit_transform(inp),
np.hstack(([[0], [0], [0]], indicator_mat)))
assert_array_equal(mlb.classes_, [4, 1, 3, 2])
# ensure fit is no-op as iterable is not consumed
inp = iter(inp)
mlb = MultiLabelBinarizer(classes=[1, 3, 2])
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
class MultihotEncoder(BaseEstimator, TransformerMixin):
"""
Wraps `MultiLabelBinarizer` in a pipeline safe transformer
Args:
sparse_output (bool): convert output to sparse matrix
"""
def __init__(self, sparse_output=False):
self.transformer = MultiLabelBinarizer()
self.sparse_output = sparse_output
def fit(self, X, y=None):
""" Fit MultiLabelBinarizer """
self.transformer.fit(X)
return self
def transform(self, X,y=None):
""" Transform MultiLabelBinarizer """
# ignore unseen label warning
with warnings.catch_warnings():
warnings.simplefilter("ignore")
X_t = self.transformer.transform(X)
if self.sparse_output:
return sparse.csr_matrix(X_t)
else:
return X_t
def test_multilabel_binarizer_unknown_class():
mlb = MultiLabelBinarizer()
y = [[1, 2]]
Y = np.array([[1, 0], [0, 1]])
w = 'unknown class(es) [0, 4] will be ignored'
matrix = assert_warns_message(UserWarning, w,
mlb.fit(y).transform, [[4, 1], [2, 0]])
assert_array_equal(matrix, Y)
Y = np.array([[1, 0, 0], [0, 1, 0]])
mlb = MultiLabelBinarizer(classes=[1, 2, 3])
matrix = assert_warns_message(UserWarning, w,
mlb.fit(y).transform, [[4, 1], [2, 0]])
assert_array_equal(matrix, Y)
def test_multilabel_binarizer_unknown_class():
mlb = MultiLabelBinarizer()
y = [[1, 2]]
Y = np.array([[1, 0], [0, 1]])
w = 'unknown class(es) [0, 4] will be ignored'
matrix = assert_warns_message(UserWarning, w,
mlb.fit(y).transform, [[4, 1], [2, 0]])
assert_array_equal(matrix, Y)
Y = np.array([[1, 0, 0], [0, 1, 0]])
mlb = MultiLabelBinarizer(classes=[1, 2, 3])
matrix = assert_warns_message(UserWarning, w,
mlb.fit(y).transform, [[4, 1], [2, 0]])
assert_array_equal(matrix, Y)
def fit_binarizers(all_values):
binarizers = {}
for f in range(len(all_values[0])):
cur_features = [context[f] for context in all_values]
# only categorical values need to be binarized, ints/floats are left as they are
if type(cur_features[0]) == str or type(cur_features[0]) == unicode:
lb = LabelBinarizer()
lb.fit(cur_features)
binarizers[f] = lb
elif type(cur_features[0]) == list:
mlb = MultiLabelBinarizer()
# default feature for unknown values
cur_features.append(tuple(("__unk__",)))
mlb.fit([tuple(x) for x in cur_features])
binarizers[f] = mlb
return binarizers
def test_sparse_output_multilabel_binarizer():
# test input as iterable of iterables
inputs = [
lambda: [(2, 3), (1, ), (1, 2)],
lambda: (set([2, 3]), set([1]), set([1, 2])),
lambda: iter([iter((2, 3)), iter(
(1, )), set([1, 2])]),
]
indicator_mat = np.array([[0, 1, 1], [1, 0, 0], [1, 1, 0]])
inverse = inputs[0]()
for sparse_output in [True, False]:
for inp in inputs:
# With fit_tranform
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit_transform(inp())
assert_equal(issparse(got), sparse_output)
if sparse_output:
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
# With fit
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit(inp()).transform(inp())
assert_equal(issparse(got), sparse_output)
if sparse_output:
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
assert_raises(ValueError, mlb.inverse_transform,
csr_matrix(np.array([[0, 1, 1], [2, 0, 0], [1, 1, 0]])))
def fit_binarizers(all_values):
binarizers = {}
for f in range(len(all_values[0])):
cur_features = [context[f] for context in all_values]
# only categorical values need to be binarized, ints/floats are left as they are
if type(cur_features[0]) == str or type(cur_features[0]) == unicode:
lb = LabelBinarizer()
lb.fit(cur_features)
binarizers[f] = lb
elif type(cur_features[0]) == list:
mlb = MultiLabelBinarizer()
# default feature for unknown values
cur_features.append(tuple(("__unk__",)))
mlb.fit([tuple(x) for x in cur_features])
binarizers[f] = mlb
return binarizers
def test_multilabel_binarizer():
# test input as iterable of iterables
inputs = [
lambda: [(2, 3), (1,), (1, 2)],
lambda: (set([2, 3]), set([1]), set([1, 2])),
lambda: iter([iter((2, 3)), iter((1,)), set([1, 2])]),
]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 1, 0]])
inverse = inputs[0]()
for inp in inputs:
# With fit_transform
mlb = MultiLabelBinarizer()
got = mlb.fit_transform(inp())
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
# With fit
mlb = MultiLabelBinarizer()
got = mlb.fit(inp()).transform(inp())
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
def test_multilabel_binarizer_non_integer_labels():
tuple_classes = np.empty(3, dtype=object)
tuple_classes[:] = [(1,), (2,), (3,)]
inputs = [
([('2', '3'), ('1',), ('1', '2')], ['1', '2', '3']),
([('b', 'c'), ('a',), ('a', 'b')], ['a', 'b', 'c']),
([((2,), (3,)), ((1,),), ((1,), (2,))], tuple_classes),
]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 1, 0]])
for inp, classes in inputs:
# fit_transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, classes)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
# fit().transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, classes)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
mlb = MultiLabelBinarizer()
assert_raises(TypeError, mlb.fit_transform, [({}), ({}, {'a': 'b'})])
def test_multilabel_binarizer():
# test input as iterable of iterables
inputs = [
lambda: [(2, 3), (1,), (1, 2)],
lambda: ({2, 3}, {1}, {1, 2}),
lambda: iter([iter((2, 3)), iter((1,)), {1, 2}]),
]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 1, 0]])
inverse = inputs[0]()
for inp in inputs:
# With fit_transform
mlb = MultiLabelBinarizer()
got = mlb.fit_transform(inp())
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
# With fit
mlb = MultiLabelBinarizer()
got = mlb.fit(inp()).transform(inp())
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
def test_multilabel_binarizer_unknown_class():
mlb = MultiLabelBinarizer()
y = [[1, 2]]
assert_raises(KeyError, mlb.fit(y).transform, [[0]])
mlb = MultiLabelBinarizer(classes=[1, 2])
assert_raises(KeyError, mlb.fit_transform, [[0]])
def test_multilabel_binarizer_non_integer_labels():
tuple_classes = np.empty(3, dtype=object)
tuple_classes[:] = [(1,), (2,), (3,)]
inputs = [
([('2', '3'), ('1',), ('1', '2')], ['1', '2', '3']),
([('b', 'c'), ('a',), ('a', 'b')], ['a', 'b', 'c']),
([((2,), (3,)), ((1,),), ((1,), (2,))], tuple_classes),
]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 1, 0]])
for inp, classes in inputs:
# fit_transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, classes)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
# fit().transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, classes)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
mlb = MultiLabelBinarizer()
assert_raises(TypeError, mlb.fit_transform, [({}), ({}, {'a': 'b'})])
def test_multilabel_binarizer_unknown_class():
mlb = MultiLabelBinarizer()
y = [[1, 2]]
assert_raises(KeyError, mlb.fit(y).transform, [[0]])
mlb = MultiLabelBinarizer(classes=[1, 2])
assert_raises(KeyError, mlb.fit_transform, [[0]])
def test_sparse_output_multilabel_binarizer():
# test input as iterable of iterables
inputs = [
lambda: [(2, 3), (1,), (1, 2)],
lambda: (set([2, 3]), set([1]), set([1, 2])),
lambda: iter([iter((2, 3)), iter((1,)), set([1, 2])]),
]
indicator_mat = np.array([[0, 1, 1], [1, 0, 0], [1, 1, 0]])
inverse = inputs[0]()
for sparse_output in [True, False]:
for inp in inputs:
# With fit_tranform
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit_transform(inp())
assert_equal(issparse(got), sparse_output)
if sparse_output:
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
# With fit
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit(inp()).transform(inp())
assert_equal(issparse(got), sparse_output)
if sparse_output:
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
assert_raises(ValueError, mlb.inverse_transform, csr_matrix(np.array([[0, 1, 1], [2, 0, 0], [1, 1, 0]])))
class MultiLabelBinarizerImpl():
def __init__(self, classes=None, sparse_output=False):
self._hyperparams = {
'classes': classes,
'sparse_output': sparse_output
}
def fit(self, X, y=None):
self._sklearn_model = SKLModel(**self._hyperparams)
if (y is not None):
self._sklearn_model.fit(X, y)
else:
self._sklearn_model.fit(X)
return self
def transform(self, X):
return self._sklearn_model.transform(X)
def test_multilabel_binarizer_same_length_sequence():
# Ensure sequences of the same length are not interpreted as a 2-d array
inp = [[1], [0], [2]]
indicator_mat = np.array([[0, 1, 0], [1, 0, 0], [0, 0, 1]])
# fit_transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
# fit().transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
def test_multilabel_binarizer_same_length_sequence():
# Ensure sequences of the same length are not interpreted as a 2-d array
inp = [[1], [0], [2]]
indicator_mat = np.array([[0, 1, 0], [1, 0, 0], [0, 0, 1]])
# fit_transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
# fit().transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
def test_sparse_output_multilabel_binarizer():
# test input as iterable of iterables
inputs = [
lambda: [(2, 3), (1,), (1, 2)],
lambda: ({2, 3}, {1}, {1, 2}),
lambda: iter([iter((2, 3)), iter((1,)), {1, 2}]),
]
indicator_mat = np.array([[0, 1, 1],
[1, 0, 0],
[1, 1, 0]])
inverse = inputs[0]()
for sparse_output in [True, False]:
for inp in inputs:
# With fit_transform
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit_transform(inp())
assert_equal(issparse(got), sparse_output)
if sparse_output:
# verify CSR assumption that indices and indptr have same dtype
assert_equal(got.indices.dtype, got.indptr.dtype)
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
# With fit
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit(inp()).transform(inp())
assert_equal(issparse(got), sparse_output)
if sparse_output:
# verify CSR assumption that indices and indptr have same dtype
assert_equal(got.indices.dtype, got.indptr.dtype)
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert_equal(mlb.inverse_transform(got), inverse)
assert_raises(ValueError, mlb.inverse_transform,
csr_matrix(np.array([[0, 1, 1],
[2, 0, 0],
[1, 1, 0]])))
def test_sparse_output_multilabel_binarizer():
# test input as iterable of iterables
inputs = [
lambda: [(2, 3), (1, ), (1, 2)],
lambda: ({2, 3}, {1}, {1, 2}),
lambda: iter([iter((2, 3)), iter((1, )), {1, 2}]),
]
indicator_mat = np.array([[0, 1, 1], [1, 0, 0], [1, 1, 0]])
inverse = inputs[0]()
for sparse_output in [True, False]:
for inp in inputs:
# With fit_transform
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit_transform(inp())
assert issparse(got) == sparse_output
if sparse_output:
# verify CSR assumption that indices and indptr have same dtype
assert got.indices.dtype == got.indptr.dtype
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert mlb.inverse_transform(got) == inverse
# With fit
mlb = MultiLabelBinarizer(sparse_output=sparse_output)
got = mlb.fit(inp()).transform(inp())
assert issparse(got) == sparse_output
if sparse_output:
# verify CSR assumption that indices and indptr have same dtype
assert got.indices.dtype == got.indptr.dtype
got = got.toarray()
assert_array_equal(indicator_mat, got)
assert_array_equal([1, 2, 3], mlb.classes_)
assert mlb.inverse_transform(got) == inverse
with pytest.raises(ValueError):
mlb.inverse_transform(
csr_matrix(np.array([[0, 1, 1], [2, 0, 0], [1, 1, 0]])))
def test_multilabel_binarizer_non_integer_labels():
tuple_classes = np.empty(3, dtype=object)
tuple_classes[:] = [(1,), (2,), (3,)]
inputs = [
([("2", "3"), ("1",), ("1", "2")], ["1", "2", "3"]),
([("b", "c"), ("a",), ("a", "b")], ["a", "b", "c"]),
([((2,), (3,)), ((1,),), ((1,), (2,))], tuple_classes),
]
indicator_mat = np.array([[0, 1, 1], [1, 0, 0], [1, 1, 0]])
for inp, classes in inputs:
# fit_transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit_transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, classes)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
# fit().transform()
mlb = MultiLabelBinarizer()
assert_array_equal(mlb.fit(inp).transform(inp), indicator_mat)
assert_array_equal(mlb.classes_, classes)
assert_array_equal(mlb.inverse_transform(indicator_mat), inp)
mlb = MultiLabelBinarizer()
assert_raises(TypeError, mlb.fit_transform, [({}), ({}, {"a": "b"})])
nltk.download('punkt')
doc2vec_model_location = 'model/doc2vec-model.bin'
classifier_model_location = 'model/classifier-model.bin'
# Use the doc2vec model created in reuters-classifier-train.py
doc2vec = Doc2Vec.load(doc2vec_model_location)
# Load the test articles and convert it to doc vectors
test_articles = [{'raw': reuters.raw(fileId), 'categories': reuters.categories(fileId)} for fileId in reuters.fileids() if fileId.startswith('test/')]
test_data = [doc2vec.infer_vector(word_tokenize(article['raw'])) for article in test_articles]
# Initialize the neural network
model=load_model(classifier_model_location)
# Make predictions
predictions = model.predict(numpy.asarray(test_data))
# Convert the prediction with gives a value between 0 and 1 to exactly 0 or 1 with a threshold
predictions[predictions<0.5] = 0
predictions[predictions>=0.5] = 1
# Convert predicted classes back to category names
labelBinarizer = MultiLabelBinarizer()
labelBinarizer.fit([reuters.categories(fileId) for fileId in reuters.fileids()])
predicted_labels = labelBinarizer.inverse_transform(predictions)
for predicted_label, test_article in zip(predicted_labels, test_articles):
print('title: {}'.format(test_article['raw'].splitlines()[0]))
print('predicted: {} - actual: {}'.format(list(predicted_label), test_article['categories']))
print('')
from nltk.tokenize import word_tokenize
from sklearn.preprocessing.label import MultiLabelBinarizer
nltk.download('reuters')
nltk.download('punkt')
google_news_word2vec_model_location = 'data/GoogleNews-vectors-negative300.bin.gz'
doc2vec_model_location = 'model/doc2vec-model.bin'
doc2vec_dimensions = 300
classifier_model_location = 'model/classifier-model.bin'
doc2vec = Doc2Vec.load(doc2vec_model_location)
# Convert the categories to one hot encoded categories
labelBinarizer = MultiLabelBinarizer()
labelBinarizer.fit([reuters.categories(fileId) for fileId in reuters.fileids()])
# Convert load the articles with their corresponding categories
train_articles = [{'raw': reuters.raw(fileId), 'categories': reuters.categories(fileId)} for fileId in reuters.fileids() if fileId.startswith('training/')]
test_articles = [{'raw': reuters.raw(fileId), 'categories': reuters.categories(fileId)} for fileId in reuters.fileids() if fileId.startswith('test/')]
shuffle(train_articles)
shuffle(test_articles)
# Convert the articles to document vectors using the doc2vec model
train_data = [doc2vec.infer_vector(word_tokenize(article['raw'])) for article in train_articles]
test_data = [doc2vec.infer_vector(word_tokenize(article['raw'])) for article in test_articles]
train_labels = labelBinarizer.transform([article['categories'] for article in train_articles])
test_labels = labelBinarizer.transform([article['categories'] for article in test_articles])
train_data, test_data, train_labels, test_labels = numpy.asarray(train_data), numpy.asarray(test_data), numpy.asarray(train_labels), numpy.asarray(test_labels)
# Initialize the neural network
