-
Notifications
You must be signed in to change notification settings - Fork 0
/
Data.py
241 lines (223 loc) · 10.5 KB
/
Data.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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
#!/usr/bin/env python
from TimeSeries import TimeSeries
from Image import Image
import numpy as np
import pickle # in Python 3.x, cpickle is considered a detail in implementing pickle, and not separately specified by users
import csv
from sklearn.feature_extraction import DictVectorizer
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.preprocessing import LabelEncoder
from sklearn.datasets import samples_generator
#import logging, logging.config
from sklearn.cross_validation import train_test_split
class Data(object):
def __init__(self, name='', logConf='logging.conf'):
self.name = name
self.X = None
self.y = None
self.Xnames = None
self.yname = None
self.xcorpora = {}
self.vocabularies = {}
self.ytype = None
self.yEncoder = None # use yEncoder.inverse_transform to recover the original category names
self.yClasses = None
#logging.config.fileConfig(logConf)
#self.logger = logging.getLogger('autoML')
def generateClassificationData(self, n_samples=100, n_features=20, n_classes=2, n_informative=5, n_redundant=0, random_state=None):
"""Generate random data suitable for classification training tasks."""
if n_informative > n_features:
n_informative = n_features
self.X, self.y = samples_generator.make_classification(n_samples, n_features,
n_informative, n_redundant, n_classes=n_classes, random_state=random_state)
def generateRegressionData(self, n_samples=100, n_features=20, n_targets=1, n_informative=5, random_state=None):
"""Generate random data suitable for regression training tasks."""
if n_informative > n_features:
n_informative = n_features
self.X, self.y = samples_generator.make_regression(n_samples, n_features,
n_informative, n_targets=n_targets, random_state=random_state)
def generateClusteringData(self, n_samples=100, n_features=20, n_clusters=5):
"""Generate random data suitable for clustering tasks."""
self.X, self.y = samples_generator.make_blobs(n_samples, n_features, n_clusters, random_state=None)
def addNoise(self, level=0.1):
"""Add noise to non-categorical variables and return a matrix. The level of the noise is specified as a
parameter. For each variable, the amount of noise is bouded by [(1-level)*value , (1+level)*value]"""
Xnew = np.copy(self.X)
for i in range(Xnew.shape[1]):
# Skip binary columns
if not set(np.unique(Xnew[:,i])) == set([0,1]):
noise = np.random.uniform(1-level, 1+level, Xnew.shape[0])
Xnew[:,i] *= noise
return Xnew
def parseCSVHeader(self, header, datatypesep="__"):
"""Each column name is expected to be of the form <name><sep><type>.
For example, height__float, age__int, race__cat represent three
variables of types float, integer and category."""
hd = [c.split(datatypesep) for c in header]
# If the user hasn't specified the data type, then assume it to be a float
for i in range(len(hd)):
if len(hd[i]) == 1:
hd[i].append('float')
return [list(t) for t in zip(*hd)]
def parseRawRow(self, row, header, types, ts_sep=';'):
"""X data is parsed and returned as a dictionary. Missing values are
stored as np.nan. The returned data can later be processed with
sklearn.feature_extraction.DictVectorizer() to handle categorical
and missing variables."""
rowDict = {}
for i, v in enumerate(row):
if len(v.strip()) > 0:
if types[i] == "int":
rowDict[header[i]] = round(float(v), 0)
elif types[i] == "float":
rowDict[header[i]] = float(v)
elif types[i] == "ts":
# seems the following is more clear: vals = [float(tsvs) for tsvs in v.split(ts_sep) if len(tsvs) > 0]
vals = [float(tsv) for tsv in [tsvs for tsvs in v.split(ts_sep) if len(tsvs) > 0]]
ts = TimeSeries(vals, 100, )
features = ts.getFeatures()
for j, f in enumerate(features):
rowDict["%s_%d" % (header[i], j)] = f
elif types[i] == 'text':
if not i in self.xcorpora:
self.xcorpora[i] = []
self.xcorpora[i].append(v)
elif types[i] == 'imgfile':
img = Image()
img.readFromFile(v)
features = img.getFeatures(10)
for j, f in enumerate(features):
rowDict["%s_%d" % (header[i], j)] = f
else:
rowDict[header[i]] = v
else:
rowDict[header[i]] = np.nan
return rowDict
def processRawData(self, xdata, ydata, header, types, stop_words='english'):
"""Instantiate self.X, self.y and related meta variables."""
print ( "Converting file to features" )
vec = DictVectorizer()
self.X = vec.fit_transform(xdata).toarray()
self.Xnames = vec.get_feature_names()
for i, corpus in self.xcorpora.items():
tfvec = TfidfVectorizer(stop_words=stop_words, vocabulary=self.vocabularies.get(i, None))
self.X = np.hstack((self.X, tfvec.fit_transform(corpus).todense()))
self.Xnames.extend(["%s_%s" % (header[i], f) for f in tfvec.get_feature_names()])
self.vocabularies[i] = tfvec.get_feature_names()
self.Xshape = self.X.shape
if ydata:
self.yname = header[-1]
self.ytype = types[-1]
if types[-1] == "cat":
le = LabelEncoder()
le.fit(ydata)
self.y = le.transform(ydata)
self.yEncoder = le
self.yClasses = le.classes_
else:
self.y = np.array(map(float, ydata))
def readRawDataFromCSV(self, fname, hasY=True, sep=",", datatypesep="__"):
"""Reads data from a CSV file"""
with open(fname, 'rU') as f:
xdata = []
ydata = []
r = csv.reader(f, delimiter=sep)
header, types = self.parseCSVHeader(next(r), datatypesep) # header=variable name for reach column, type=data type
for row in r:
if hasY:
xdata.append(self.parseRawRow(row[:-1], header, types)) # returns a row dictionary
ydata.append(row[-1])
else:
xdata.append(self.parseRawRow(row, header, types)) # returns a row dictionary
self.processRawData(xdata, ydata, header, types)
self.types = types
self.xdata = xdata
self.header = header
def addRows(self, other):
"""Add rows using data from a different instance of Data()."""
self.X = np.vstack((self.X, other.X))
self.y = np.vstack((self.y, other.y))
def addCols(self, other):
"""Add columns using data from a different instance of Data()."""
self.X = np.hstack((self.X, other.X))
for name in other.Xnames:
self.Xnames.append( name )
for vocab in other.vocabularies:
self.vocabularies.update( vocab )
def dropCols(self, indices):
"""Drop columns specified by the given indices."""
self.X = np.delete(self.X, indices, axis=1)
for i in sorted(indices, reverse=True):
del self.Xnames[i]
def saveToFile(self, fname):
"""Save the dataset as a Python pickle file."""
pickle.dump({'name': self.name, 'X': self.X, 'y': self.y, 'Xnames': self.Xnames, 'yname': self.yname,
'ytype': self.ytype, 'yEncoder': self.yEncoder, 'yClasses': self.yClasses},
open(fname, "wb"))
def exportToCSV(self, fname):
"""Save the dataset as a comma separated text file."""
with open(fname, 'wb') as f:
cf = csv.writer(f)
if self.y is None:
cf.writerow(self.Xnames)
else:
cf.writerow(self.Xnames + [self.yname])
for i, row in enumerate(self.X):
if self.y is None:
cf.writerow(row)
else:
cf.writerow(np.hstack((row, self.y[i])))
def loadFromFile(self, fname):
d = pickle.load(open(fname, "rb"))
self.name = d['name']
self.X = d['X']
self.y = d['y']
self.Xnames = d['Xnames']
self.yname = d['yname']
self.ytype = d['ytype']
self.yEncoder = d['yEncoder']
self.yClasses = d['yClasses']
def getTrainAndTestData(self, test_size=0.45, random_state = 1):
if self.y is not None:
return train_test_split(self.X, self.y, test_size=test_size, random_state = random_state)
else:
return train_test_split(self.X, test_size=test_size, random_state = random_state)
def __str__(self):
if self.y is None and self.X is not None:
return(
"""Dataset '%s': %s
Column names: %s
Row 1: %s
Row -1: %s""" % (self.name, self.X.shape, self.Xnames, self.X[0, :], self.X[-1, :]))
elif self.X is not None:
return(
"""Dataset '%s': %s
Column names: %s
Target name: %s
Target type: %s
Target classes: %s
Target encoding: %s
Row 1: %s -> %s
Row -1: %s -> %s""" % (self.name, self.X.shape, self.Xnames, self.yname, self.ytype, self.yClasses,
(self.yEncoder.transform(self.yClasses) if self.yEncoder else ''),
self.X[0, :], self.y[0], self.X[-1, :], self.y[-1]))
else:
return("Empty Dataset '%s'" % self.name)
#########
if __name__ == '__main__':
#d = Data("test")
#d.readRawDataFromCSV("test.csv")
#print d
#d.exportToCSV('regtest.csv')
g = Data("classification")
g.generateClassificationData(n_features=2, n_samples=10)
print ( g )
print ( g.addNoise() )
#h = Data("regression")
#h.generateRegressionData()
#print h
#d = Data('comr.se')
#d.readRawDataFromHive(host='13.4.40.80', username='bigdatafoundry', password=None,
# database='bdf_prd_comrse_prd', table='transaction', maxRows=1000,
# useCols=frozenset([2,3,4,5,9,11,12,13]))
#print d