def __init__(self, lemmatization=False):
BugModel.__init__(self, lemmatization)
self.sampler = BorderlineSMOTE(random_state=0)
self.calculate_importance = False
feature_extractors = [
bug_features.has_str(),
bug_features.has_regression_range(),
bug_features.severity(),
bug_features.has_crash_signature(),
bug_features.has_url(),
bug_features.whiteboard(),
bug_features.product(),
# TODO: We would like to use the component at the time of filing too,
# but we can't because the rollback script doesn't support changes to
# components yet.
# bug_features.component(),
bug_features.num_words_title(),
bug_features.num_words_comments(),
bug_features.keywords(),
bug_features.priority(),
bug_features.version(),
bug_features.target_milestone(),
bug_features.has_attachment(),
bug_features.platform(),
bug_features.op_sys(),
]
cleanup_functions = [
feature_cleanup.fileref(),
feature_cleanup.url(),
feature_cleanup.synonyms(),
]
self.extraction_pipeline = Pipeline([
(
"bug_extractor",
bug_features.BugExtractor(feature_extractors,
cleanup_functions,
rollback=True),
),
(
"union",
ColumnTransformer([
("data", DictVectorizer(), "data"),
("title", self.text_vectorizer(min_df=0.0001), "title"),
(
"comments",
self.text_vectorizer(min_df=0.0001),
"comments",
),
]),
),
])
self.clf = xgboost.XGBClassifier(n_jobs=utils.get_physical_cpu_count())
self.clf.set_params(predictor="cpu_predictor")
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.short_desc_maxlen = 20
self.short_desc_vocab_size = 25000
self.short_desc_emb_sz = 300
self.long_desc_maxlen = 100
self.long_desc_vocab_size = 25000
self.long_desc_emb_sz = 300
self.cross_validation_enabled = False
self.params = [
{
"short_desc_emb_dropout_rate": 0.2,
"short_desc_encoded_gru_units": 256,
"short_desc_encoded_gru_dropout": 0.45,
"short_desc_encoded_recurrent_dropout": 0.5,
"long_desc_emb_dropout_rate": 0.25,
"long_desc_encoded_gru_units": 256,
"long_desc_encoded_dropout": 0.5,
"long_desc_encoded_recurrent_dropout": 0.55,
"rep_platform_emb_input_dim": 14,
"rep_platform_emb_output_dim": 25,
"rep_platform_emb_spatial_dropout_rate": 0.1,
"rep_platform_emb_dropout_rate": 0.45,
"op_sys_emb_input_dim": 48,
"op_sys_emb_output_dim": 50,
"op_sys_emb_spatial_dropout_rate": 0.1,
"op_sys_emb_dropout_rate": 0.45,
"reporter_emb_input_dim": 46544,
"reporter_emb_output_dim": 100,
"reporter_emb_spatial_dropout_rate": 0.15,
"reporter_emb_dropout_rate": 0.5,
"tfidf_word_dense_units": 600,
"tfidf_word_dropout_rate": 0.5,
"tfidf_char_inp_dense_unit": 500,
"tfidf_char_inp_dropout_rate": 0.5,
"x_dense_unit": 2000,
"x_dropout_rate": 0.6,
},
{
"short_desc_emb_dropout_rate": 0.2,
"short_desc_encoded_gru_units": 250,
"short_desc_encoded_gru_dropout": 0.45,
"short_desc_encoded_recurrent_dropout": 0.45,
"long_desc_emb_dropout_rate": 0.25,
"long_desc_encoded_gru_units": 250,
"long_desc_encoded_dropout": 0.45,
"long_desc_encoded_recurrent_dropout": 0.45,
"rep_platform_emb_input_dim": 14,
"rep_platform_emb_output_dim": 30,
"rep_platform_emb_spatial_dropout_rate": 0.1,
"rep_platform_emb_dropout_rate": 0.4,
"op_sys_emb_input_dim": 48,
"op_sys_emb_output_dim": 55,
"op_sys_emb_spatial_dropout_rate": 0.1,
"op_sys_emb_dropout_rate": 0.4,
"reporter_emb_input_dim": 46544,
"reporter_emb_output_dim": 110,
"reporter_emb_spatial_dropout_rate": 0.15,
"reporter_emb_dropout_rate": 0.45,
"tfidf_word_dense_units": 610,
"tfidf_word_dropout_rate": 0.45,
"tfidf_char_inp_dense_unit": 510,
"tfidf_char_inp_dropout_rate": 0.5,
"x_dense_unit": 1970,
"x_dropout_rate": 0.5,
},
{
"short_desc_emb_dropout_rate": 0.2,
"short_desc_encoded_gru_units": 266,
"short_desc_encoded_gru_dropout": 0.45,
"short_desc_encoded_recurrent_dropout": 0.45,
"long_desc_emb_dropout_rate": 0.25,
"long_desc_encoded_gru_units": 266,
"long_desc_encoded_dropout": 0.45,
"long_desc_encoded_recurrent_dropout": 0.55,
"rep_platform_emb_input_dim": 14,
"rep_platform_emb_output_dim": 35,
"rep_platform_emb_spatial_dropout_rate": 0.1,
"rep_platform_emb_dropout_rate": 0.45,
"op_sys_emb_input_dim": 48,
"op_sys_emb_output_dim": 60,
"op_sys_emb_spatial_dropout_rate": 0.1,
"op_sys_emb_dropout_rate": 0.45,
"reporter_emb_input_dim": 46544,
"reporter_emb_output_dim": 120,
"reporter_emb_spatial_dropout_rate": 0.15,
"reporter_emb_dropout_rate": 0.45,
"tfidf_word_dense_units": 620,
"tfidf_word_dropout_rate": 0.5,
"tfidf_char_inp_dense_unit": 520,
"tfidf_char_inp_dropout_rate": 0.45,
"x_dense_unit": 1950,
"x_dropout_rate": 0.5,
},
]
feature_extractors = [
bug_features.bug_reporter(),
bug_features.platform(),
bug_features.op_sys(),
]
cleanup_functions = []
self.extraction_pipeline = Pipeline([
(
"bug_extractor",
bug_features.BugExtractor(feature_extractors,
cleanup_functions),
),
(
"union",
StructuredColumnTransformer([
(
"platform",
make_pipeline(DictExtractor("platform"),
OrdinalEncoder()),
"data",
),
(
"op_sys",
make_pipeline(DictExtractor("op_sys"),
OrdinalEncoder()),
"data",
),
(
"bug_reporter",
make_pipeline(
DictExtractor("bug_reporter"),
MissingOrdinalEncoder(),
),
"data",
),
(
"title_sequence",
KerasTextToSequences(self.short_desc_maxlen,
self.short_desc_vocab_size),
"title",
),
(
"first_comment_sequence",
KerasTextToSequences(self.long_desc_maxlen,
self.long_desc_vocab_size),
"first_comment",
),
(
"title_char_tfidf",
TfidfVectorizer(
strip_accents="unicode",
analyzer="char",
stop_words="english",
ngram_range=(2, 4),
max_features=25000,
sublinear_tf=True,
),
"title",
),
(
"title_word_tfidf",
TfidfVectorizer(
strip_accents="unicode",
min_df=0.0001,
max_df=0.1,
analyzer="word",
token_pattern=r"\w{1,}",
stop_words="english",
ngram_range=(2, 4),
max_features=30000,
sublinear_tf=True,
),
"title",
),
]),
),
])
kwargs = {
"short_desc_maxlen": self.short_desc_maxlen,
"short_desc_vocab_size": self.short_desc_vocab_size,
"short_desc_emb_sz": self.short_desc_emb_sz,
"long_desc_maxlen": self.long_desc_maxlen,
"long_desc_vocab_size": self.long_desc_vocab_size,
"long_desc_emb_sz": self.long_desc_emb_sz,
}
estimators = []
for i, params in enumerate(self.params):
kwargs["params"] = params
estimator = ComponentNNClassifier(**kwargs)
estimators.append(("model_{}".format(i), estimator))
self.clf = VotingClassifier(estimators=estimators,
voting="soft",
weights=[1, 1, 1])
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.short_desc_maxlen = 20
self.short_desc_vocab_size = 25000
self.short_desc_emb_sz = 300
self.long_desc_maxlen = 100
self.long_desc_vocab_size = 25000
self.long_desc_emb_sz = 300
self.cross_validation_enabled = False
self.params = [{
'short_desc_emb_dropout_rate': 0.2,
'short_desc_encoded_gru_units': 256,
'short_desc_encoded_gru_dropout': 0.45,
'short_desc_encoded_recurrent_dropout': 0.5,
'long_desc_emb_dropout_rate': 0.25,
'long_desc_encoded_gru_units': 256,
'long_desc_encoded_dropout': 0.5,
'long_desc_encoded_recurrent_dropout': 0.55,
'rep_platform_emb_input_dim': 14,
'rep_platform_emb_output_dim': 25,
'rep_platform_emb_spatial_dropout_rate': 0.1,
'rep_platform_emb_dropout_rate': 0.45,
'op_sys_emb_input_dim': 48,
'op_sys_emb_output_dim': 50,
'op_sys_emb_spatial_dropout_rate': 0.1,
'op_sys_emb_dropout_rate': 0.45,
'reporter_emb_input_dim': 46544,
'reporter_emb_output_dim': 100,
'reporter_emb_spatial_dropout_rate': 0.15,
'reporter_emb_dropout_rate': 0.5,
'tfidf_word_dense_units': 600,
'tfidf_word_dropout_rate': 0.5,
'tfidf_char_inp_dense_unit': 500,
'tfidf_char_inp_dropout_rate': 0.5,
'x_dense_unit': 2000,
'x_dropout_rate': 0.6,
}, {
'short_desc_emb_dropout_rate': 0.2,
'short_desc_encoded_gru_units': 250,
'short_desc_encoded_gru_dropout': 0.45,
'short_desc_encoded_recurrent_dropout': 0.45,
'long_desc_emb_dropout_rate': 0.25,
'long_desc_encoded_gru_units': 250,
'long_desc_encoded_dropout': 0.45,
'long_desc_encoded_recurrent_dropout': 0.45,
'rep_platform_emb_input_dim': 14,
'rep_platform_emb_output_dim': 30,
'rep_platform_emb_spatial_dropout_rate': 0.1,
'rep_platform_emb_dropout_rate': 0.4,
'op_sys_emb_input_dim': 48,
'op_sys_emb_output_dim': 55,
'op_sys_emb_spatial_dropout_rate': 0.1,
'op_sys_emb_dropout_rate': 0.4,
'reporter_emb_input_dim': 46544,
'reporter_emb_output_dim': 110,
'reporter_emb_spatial_dropout_rate': 0.15,
'reporter_emb_dropout_rate': 0.45,
'tfidf_word_dense_units': 610,
'tfidf_word_dropout_rate': 0.45,
'tfidf_char_inp_dense_unit': 510,
'tfidf_char_inp_dropout_rate': 0.5,
'x_dense_unit': 1970,
'x_dropout_rate': 0.5,
}, {
'short_desc_emb_dropout_rate': 0.2,
'short_desc_encoded_gru_units': 266,
'short_desc_encoded_gru_dropout': 0.45,
'short_desc_encoded_recurrent_dropout': 0.45,
'long_desc_emb_dropout_rate': 0.25,
'long_desc_encoded_gru_units': 266,
'long_desc_encoded_dropout': 0.45,
'long_desc_encoded_recurrent_dropout': 0.55,
'rep_platform_emb_input_dim': 14,
'rep_platform_emb_output_dim': 35,
'rep_platform_emb_spatial_dropout_rate': 0.1,
'rep_platform_emb_dropout_rate': 0.45,
'op_sys_emb_input_dim': 48,
'op_sys_emb_output_dim': 60,
'op_sys_emb_spatial_dropout_rate': 0.1,
'op_sys_emb_dropout_rate': 0.45,
'reporter_emb_input_dim': 46544,
'reporter_emb_output_dim': 120,
'reporter_emb_spatial_dropout_rate': 0.15,
'reporter_emb_dropout_rate': 0.45,
'tfidf_word_dense_units': 620,
'tfidf_word_dropout_rate': 0.5,
'tfidf_char_inp_dense_unit': 520,
'tfidf_char_inp_dropout_rate': 0.45,
'x_dense_unit': 1950,
'x_dropout_rate': 0.5,
}]
feature_extractors = [
bug_features.bug_reporter(),
bug_features.platform(),
bug_features.op_sys()
]
cleanup_functions = []
self.extraction_pipeline = Pipeline([
('bug_extractor',
bug_features.BugExtractor(feature_extractors, cleanup_functions)),
('union',
StructuredColumnTransformer([
('platform',
make_pipeline(DictExtractor('platform'),
OrdinalEncoder()), 'data'),
('op_sys',
make_pipeline(DictExtractor('op_sys'),
OrdinalEncoder()), 'data'),
('bug_reporter',
make_pipeline(DictExtractor('bug_reporter'),
MissingOrdinalEncoder()), 'data'),
('title_sequence',
KerasTextToSequences(self.short_desc_maxlen,
self.short_desc_vocab_size), 'title'),
('first_comment_sequence',
KerasTextToSequences(self.long_desc_maxlen,
self.long_desc_vocab_size),
'first_comment'),
('title_char_tfidf',
TfidfVectorizer(strip_accents='unicode',
analyzer='char',
stop_words='english',
ngram_range=(2, 4),
max_features=25000,
sublinear_tf=True), 'title'),
('title_word_tfidf',
TfidfVectorizer(strip_accents='unicode',
min_df=0.0001,
max_df=0.1,
analyzer='word',
token_pattern=r'\w{1,}',
stop_words='english',
ngram_range=(2, 4),
max_features=30000,
sublinear_tf=True), 'title')
])),
])
kwargs = {
'short_desc_maxlen': self.short_desc_maxlen,
'short_desc_vocab_size': self.short_desc_vocab_size,
'short_desc_emb_sz': self.short_desc_emb_sz,
'long_desc_maxlen': self.long_desc_maxlen,
'long_desc_vocab_size': self.long_desc_vocab_size,
'long_desc_emb_sz': self.long_desc_emb_sz
}
estimators = []
for i, params in enumerate(self.params):
kwargs['params'] = params
estimator = ComponentNNClassifier(**kwargs)
estimators.append(('model_{}'.format(i), estimator))
self.clf = VotingClassifier(estimators=estimators,
voting='soft',
weights=[1, 1, 1])
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.short_desc_maxlen = 20
self.short_desc_vocab_size = 25000
self.short_desc_emb_sz = 300
self.long_desc_maxlen = 100
self.long_desc_vocab_size = 25000
self.long_desc_emb_sz = 300
self.cross_validation_enabled = False
feature_extractors = [
bug_features.bug_reporter(),
bug_features.platform(),
bug_features.op_sys()
]
cleanup_functions = []
self.extraction_pipeline = Pipeline([
('bug_extractor',
bug_features.BugExtractor(feature_extractors, cleanup_functions)),
('union',
StructuredColumnTransformer([
('platform',
make_pipeline(DictExtractor('platform'),
OrdinalEncoder()), 'data'),
('op_sys',
make_pipeline(DictExtractor('op_sys'),
OrdinalEncoder()), 'data'),
('bug_reporter',
make_pipeline(DictExtractor('bug_reporter'),
OrdinalEncoder()), 'data'),
('title_sequence',
KerasTextToSequences(self.short_desc_maxlen,
self.short_desc_vocab_size), 'title'),
('first_comment_sequence',
KerasTextToSequences(self.long_desc_maxlen,
self.long_desc_vocab_size),
'first_comment'),
('title_char_tfidf',
TfidfVectorizer(strip_accents='unicode',
analyzer='char',
stop_words='english',
ngram_range=(2, 4),
max_features=25000,
sublinear_tf=True), 'title'),
('title_word_tfidf',
TfidfVectorizer(strip_accents='unicode',
min_df=0.0001,
max_df=0.1,
analyzer='word',
token_pattern=r'\w{1,}',
stop_words='english',
ngram_range=(2, 4),
max_features=30000,
sublinear_tf=True), 'title')
])),
])
kwargs = {
'short_desc_maxlen': self.short_desc_maxlen,
'short_desc_vocab_size': self.short_desc_vocab_size,
'short_desc_emb_sz': self.short_desc_emb_sz,
'long_desc_maxlen': self.long_desc_maxlen,
'long_desc_vocab_size': self.long_desc_vocab_size,
'long_desc_emb_sz': self.long_desc_emb_sz
}
self.clf = ComponentNNClassifier(**kwargs)
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.short_desc_maxlen = 20
self.short_desc_vocab_size = 25000
self.short_desc_emb_sz = 300
self.long_desc_maxlen = 100
self.long_desc_vocab_size = 25000
self.long_desc_emb_sz = 300
self.cross_validation_enabled = False
self.params = [
{
"short_desc_emb_dropout_rate": 0.2,
"short_desc_encoded_gru_units": 256,
"short_desc_encoded_gru_dropout": 0.45,
"short_desc_encoded_recurrent_dropout": 0.5,
"long_desc_emb_dropout_rate": 0.25,
"long_desc_encoded_gru_units": 256,
"long_desc_encoded_dropout": 0.5,
"long_desc_encoded_recurrent_dropout": 0.55,
"rep_platform_emb_input_dim": 14,
"rep_platform_emb_output_dim": 25,
"rep_platform_emb_spatial_dropout_rate": 0.1,
"rep_platform_emb_dropout_rate": 0.45,
"op_sys_emb_input_dim": 48,
"op_sys_emb_output_dim": 50,
"op_sys_emb_spatial_dropout_rate": 0.1,
"op_sys_emb_dropout_rate": 0.45,
"reporter_emb_input_dim": 46544,
"reporter_emb_output_dim": 100,
"reporter_emb_spatial_dropout_rate": 0.15,
"reporter_emb_dropout_rate": 0.5,
"tfidf_word_dense_units": 600,
"tfidf_word_dropout_rate": 0.5,
"tfidf_char_inp_dense_unit": 500,
"tfidf_char_inp_dropout_rate": 0.5,
"x_dense_unit": 2000,
"x_dropout_rate": 0.6,
},
{
"short_desc_emb_dropout_rate": 0.2,
"short_desc_encoded_gru_units": 250,
"short_desc_encoded_gru_dropout": 0.45,
"short_desc_encoded_recurrent_dropout": 0.45,
"long_desc_emb_dropout_rate": 0.25,
"long_desc_encoded_gru_units": 250,
"long_desc_encoded_dropout": 0.45,
"long_desc_encoded_recurrent_dropout": 0.45,
"rep_platform_emb_input_dim": 14,
"rep_platform_emb_output_dim": 30,
"rep_platform_emb_spatial_dropout_rate": 0.1,
"rep_platform_emb_dropout_rate": 0.4,
"op_sys_emb_input_dim": 48,
"op_sys_emb_output_dim": 55,
"op_sys_emb_spatial_dropout_rate": 0.1,
"op_sys_emb_dropout_rate": 0.4,
"reporter_emb_input_dim": 46544,
"reporter_emb_output_dim": 110,
"reporter_emb_spatial_dropout_rate": 0.15,
"reporter_emb_dropout_rate": 0.45,
"tfidf_word_dense_units": 610,
"tfidf_word_dropout_rate": 0.45,
"tfidf_char_inp_dense_unit": 510,
"tfidf_char_inp_dropout_rate": 0.5,
"x_dense_unit": 1970,
"x_dropout_rate": 0.5,
},
{
"short_desc_emb_dropout_rate": 0.2,
"short_desc_encoded_gru_units": 266,
"short_desc_encoded_gru_dropout": 0.45,
"short_desc_encoded_recurrent_dropout": 0.45,
"long_desc_emb_dropout_rate": 0.25,
"long_desc_encoded_gru_units": 266,
"long_desc_encoded_dropout": 0.45,
"long_desc_encoded_recurrent_dropout": 0.55,
"rep_platform_emb_input_dim": 14,
"rep_platform_emb_output_dim": 35,
"rep_platform_emb_spatial_dropout_rate": 0.1,
"rep_platform_emb_dropout_rate": 0.45,
"op_sys_emb_input_dim": 48,
"op_sys_emb_output_dim": 60,
"op_sys_emb_spatial_dropout_rate": 0.1,
"op_sys_emb_dropout_rate": 0.45,
"reporter_emb_input_dim": 46544,
"reporter_emb_output_dim": 120,
"reporter_emb_spatial_dropout_rate": 0.15,
"reporter_emb_dropout_rate": 0.45,
"tfidf_word_dense_units": 620,
"tfidf_word_dropout_rate": 0.5,
"tfidf_char_inp_dense_unit": 520,
"tfidf_char_inp_dropout_rate": 0.45,
"x_dense_unit": 1950,
"x_dropout_rate": 0.5,
},
]
feature_extractors = [
bug_features.bug_reporter(),
bug_features.platform(),
bug_features.op_sys(),
]
cleanup_functions = []
self.extraction_pipeline = Pipeline(
[
(
"bug_extractor",
bug_features.BugExtractor(feature_extractors, cleanup_functions),
),
(
"union",
StructuredColumnTransformer(
[
(
"platform",
make_pipeline(
DictExtractor("platform"), OrdinalEncoder()
),
"data",
),
(
"op_sys",
make_pipeline(
DictExtractor("op_sys"), OrdinalEncoder()
),
"data",
),
(
"bug_reporter",
make_pipeline(
DictExtractor("bug_reporter"),
MissingOrdinalEncoder(),
),
"data",
),
(
"title_sequence",
KerasTextToSequences(
self.short_desc_maxlen, self.short_desc_vocab_size
),
"title",
),
(
"first_comment_sequence",
KerasTextToSequences(
self.long_desc_maxlen, self.long_desc_vocab_size
),
"first_comment",
),
(
"title_char_tfidf",
TfidfVectorizer(
strip_accents="unicode",
analyzer="char",
stop_words="english",
ngram_range=(2, 4),
max_features=25000,
sublinear_tf=True,
),
"title",
),
(
"title_word_tfidf",
TfidfVectorizer(
strip_accents="unicode",
min_df=0.0001,
max_df=0.1,
analyzer="word",
token_pattern=r"\w{1,}",
stop_words="english",
ngram_range=(2, 4),
max_features=30000,
sublinear_tf=True,
),
"title",
),
]
),
),
]
)
kwargs = {
"short_desc_maxlen": self.short_desc_maxlen,
"short_desc_vocab_size": self.short_desc_vocab_size,
"short_desc_emb_sz": self.short_desc_emb_sz,
"long_desc_maxlen": self.long_desc_maxlen,
"long_desc_vocab_size": self.long_desc_vocab_size,
"long_desc_emb_sz": self.long_desc_emb_sz,
}
estimators = []
for i, params in enumerate(self.params):
kwargs["params"] = params
estimator = ComponentNNClassifier(**kwargs)
estimators.append(("model_{}".format(i), estimator))
self.clf = VotingClassifier(
estimators=estimators, voting="soft", weights=[1, 1, 1]
)