/
pipelines.py
186 lines (152 loc) · 7.67 KB
/
pipelines.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
import unittest
import mock
from pyspark import SparkContext, SQLContext
from pyspark.ml import Pipeline, PipelineModel
from pyspark.ml.classification import LogisticRegression, RandomForestClassifier, MultilayerPerceptronClassifier
from pyspark.ml.feature import StringIndexer, VectorAssembler, PCA, VectorIndexer, Normalizer, Word2Vec, \
StopWordsRemover
from pyspark.ml.evaluation import BinaryClassificationEvaluator
from pyspark.ml.feature import HashingTF, Tokenizer, IDF, NGram
from pyspark.ml.tuning import ParamGridBuilder, CrossValidator
from pyspark.mllib.evaluation import BinaryClassificationMetrics
from pyspark.mllib.linalg import Vectors
from datasources import Datasources
from transformers import BeautifulSoupParser, Tokenizzzer, PorterStemmerTransformer, VaderPolarizer
from pyspark.ml.util import keyword_only
from nltk.tokenize import TweetTokenizer, WhitespaceTokenizer
class PipelineEngine(object):
def __init__(self, cv):
self.cv = cv
self.evaluator = BinaryClassificationEvaluator()
self.pipeline = None
self.param_grid = None
self.model = None
def _build_stages(self):
raise NotImplementedError()
def _build_param_grid(self):
raise NotImplementedError()
def fit(self, train):
"""
Train and return a model on the dataframe
Args:
train (Dataframe):
"""
self.cv.setEstimator(self.pipeline)
self.cv.setEvaluator(self.evaluator)
self.cv.setEstimatorParamMaps(self.param_grid)
self.model = self.cv.fit(train)
return self.model
def evaluate(self, train):
train, test = train.randomSplit([0.6, 0.4], 1234)
model = self.fit(train)
prediction = model.transform(test)
prediction_and_labels = prediction.rdd.map(lambda s: (s.prediction, s.label))
# print("Params map: " + str(self.cv.metrics))
# print("Metrics: " + str(self.cv.getEstimatorParamMaps()))
return BinaryClassificationMetrics(prediction_and_labels)
class BaselinePipelineEngine(PipelineEngine):
@keyword_only
def __init__(self, cv):
super(BaselinePipelineEngine, self).__init__(cv)
self.hashing_tf_map = [pow(2, 20)]
self.lr_map = [0.1, 0.01]
self.stages = self._build_stages()
self.pipeline = Pipeline(stages=[self.bs_parser, self.tokenizer, self.hashing_tf, self.idf_model, self.lr])
self.param_grid = self._build_param_grid()
def _build_stages(self):
self.bs_parser = BeautifulSoupParser(inputCol="review", outputCol="parsed")
self.tokenizer = Tokenizer(inputCol=self.bs_parser.getOutputCol(), outputCol="words")
self.hashing_tf = HashingTF(inputCol=self.tokenizer.getOutputCol(), outputCol="raw_features")
self.idf_model = IDF(inputCol=self.hashing_tf.getOutputCol(), outputCol="features")
self.lr = LogisticRegression(maxIter=10, regParam=0.01)
return [self.bs_parser, self.tokenizer, self.hashing_tf, self.idf_model, self.lr]
def _build_param_grid(self):
param_grid_builder = ParamGridBuilder()
param_grid_builder.addGrid(self.hashing_tf.numFeatures, self.hashing_tf_map)
param_grid_builder.addGrid(self.lr.regParam, self.lr_map)
return param_grid_builder.build()
class SentimentalPipelineEngine(PipelineEngine):
def __init__(self, cv):
super(SentimentalPipelineEngine, self).__init__(cv)
self.tokenizer_map = [TweetTokenizer()]
self.ngram_map = [1]
self.hashing_tf_map = [pow(2, 20)]
self.clf_map = [0.1]
self.stages = self._build_stages()
self.pipeline = Pipeline(stages=self.stages)
self.param_grid = self._build_param_grid()
def _build_stages(self):
self.bs_parser = BeautifulSoupParser(inputCol="review", outputCol="parsed")
self.tokenizer = Tokenizzzer(inputCol=self.bs_parser.getOutputCol(), outputCol="words")
self.stopwords_remover = StopWordsRemover(inputCol="words", outputCol="filtered")
self.porter = PorterStemmerTransformer(inputCol=self.stopwords_remover.getOutputCol(), outputCol="stemmed")
self.ngram = NGram(inputCol=self.porter.getOutputCol(), outputCol="ngrams")
self.hashing_tf = HashingTF(inputCol=self.ngram.getOutputCol(), outputCol="features")
self.idf = IDF(inputCol="features", outputCol="idf_features")
self.normalizer = Normalizer(inputCol="idf_features", outputCol="norm_features", p=1.0)
self.clf = LogisticRegression(featuresCol='norm_features', regParam=0.1)
# self.clf = MultilayerPerceptronClassifier(featuresCol="norm_features", maxIter=1000, layers=[self.hashing_tf.getNumFeatures(), 200, 100, 2])
return [self.bs_parser, self.tokenizer, self.stopwords_remover, self.porter, self.ngram, self.hashing_tf, self.idf, self.normalizer, self.clf]
def _build_param_grid(self):
param_grid_builder = ParamGridBuilder()
param_grid_builder.addGrid(self.tokenizer.tokenizer, self.tokenizer_map)
param_grid_builder.addGrid(self.ngram.n, self.ngram_map)
param_grid_builder.addGrid(self.hashing_tf.numFeatures, self.hashing_tf_map)
param_grid_builder.addGrid(self.clf.regParam, self.clf_map)
return param_grid_builder.build()
class TestPipeline(PipelineEngine):
def __init__(self, cv):
super(TestPipeline, self).__init__(cv)
self.lr_map = [0.01]
self.stages = self._build_stages()
self.pipeline = Pipeline(stages=[self.lr])
self.param_grid = self._build_param_grid()
def _build_stages(self):
self.lr = LogisticRegression(maxIter=10, regParam=0.01)
return [self.lr]
def _build_param_grid(self):
param_grid_builder = ParamGridBuilder()
param_grid_builder.addGrid(self.lr.regParam, self.lr_map)
return param_grid_builder.build()
class SparkTest(unittest.TestCase):
def setUp(self):
self.sc = SparkContext("local", "test_app")
def tearDown(self):
self.sc.stop()
class SentimentalPipelineTest(SparkTest):
def test_build_stages(self):
self.pipeline = SentimentalPipelineEngine(cv=CrossValidator())
self.assertEqual(len(self.pipeline.stages), 6)
def test_build_param_grid(self):
self.pipeline = SentimentalPipelineEngine(cv=CrossValidator())
param_grid = self.pipeline.param_grid
self.assertEqual(len(param_grid),
len(self.pipeline.ngram_map) * len(self.pipeline.hashing_tf_map) * len(self.pipeline.tokenizer_map)
)
class BaselinePipelineTest(SparkTest):
def test_build_stages(self):
self.pipeline = BaselinePipelineEngine(cv=CrossValidator())
self.assertEqual(len(self.pipeline.stages), 5)
def test_build_param_grid(self):
self.pipeline = BaselinePipelineEngine(cv=CrossValidator())
param_grid = self.pipeline.param_grid
self.assertEqual(len(param_grid), len(self.pipeline.lr_map) * len(self.pipeline.hashing_tf_map))
class PipelineEngineTest(SparkTest):
@mock.patch('pyspark.ml.tuning.CrossValidator')
def test_fit(self, mock_cv):
mock_cv.fit.return_value = PipelineModel(stages=[])
test_pipeline = TestPipeline(mock_cv)
train = self._get_train_data()
test_pipeline.fit(train)
mock_cv.fit.assert_called_with(train)
def _get_train_data(self):
sql_context = SQLContext(self.sc)
l = [
(1, Vectors.dense([1, 2, 3]), 1.0),
(2, Vectors.dense([1, 2, 3]), 0.0),
(3, Vectors.dense([1, 2, 3]), 1.0),
(4, Vectors.dense([1, 2, 3]), 0.0),
]
return sql_context.createDataFrame(l, ['id', 'features', 'label'])
if __name__ == "__main__":
unittest.main()