/
predictors.py
416 lines (344 loc) · 15.2 KB
/
predictors.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
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
# -*- coding: utf-8 -*-
"""
Created on Fri Nov 13 13:48:25 2015
@author: abzooba
"""
import os
import pandas as pd
import numpy as np
from sklearn.cross_validation import StratifiedKFold
#from sklearn.metrics import confusion_matrix
import xgboost as xgb
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import LabelEncoder
class Classifier(object):
def __init__(self, y):
self.y = y
self.classifier = None
self.noOfClasses = len( y.unique() )
if self.noOfClasses > 2:
self.classifier = MulticlassClassifier(y)
else:
self.classifier = BinaryClassifier(y)
def model(self):
self.classifier.model()
class Regressor(object):
def __init__(self, y, featureMixer, id_col, targets, eval_metric):
self.directory = featureMixer.directory
self.id_col = id_col
self.y = y
self.train = featureMixer.train
self.test = featureMixer.test
self.features = featureMixer.features
self.targets = targets
self.eval_metric = eval_metric
def mape( actual, predicted ):
return np.mean(np.abs((actual - predicted) / actual))
def sle(self, actual, predicted):
"""
Computes the squared log error.
This function computes the squared log error between two numbers,
or for element between a pair of lists or numpy arrays.
Parameters
----------
actual : int, float, list of numbers, numpy array
The ground truth value
predicted : same type as actual
The predicted value
Returns
-------
score : double or list of doubles
The squared log error between actual and predicted
"""
_sle_ = np.power( np.log( np.array(actual)+1 ) - np.log( np.array(predicted)+1 ), 2)
return _sle_
def se(self, actual, predicted):
"""
Computes the squared error.
This function computes the squared error between two numbers,
or for element between a pair of lists or numpy arrays.
Parameters
----------
actual : int, float, list of numbers, numpy array
The ground truth value
predicted : same type as actual
The predicted value
Returns
-------
score : double or list of doubles
The squared error between actual and predicted
"""
return np.power(np.array(actual)-np.array(predicted), 2)
def msle(self, actual, predicted):
"""
Computes the mean squared log error.
This function computes the mean squared log error between two lists
of numbers.
Parameters
----------
actual : list of numbers, numpy array
The ground truth value
predicted : same type as actual
The predicted value
Returns
-------
score : double
The mean squared log error between actual and predicted
"""
_msle_ = np.nanmean(self.sle(actual, predicted))
return _msle_
def mse(self, actual, predicted):
"""
Computes the mean squared error.
This function computes the mean squared error between two lists
of numbers.
Parameters
----------
actual : list of numbers, numpy array
The ground truth value
predicted : same type as actual
The predicted value
Returns
-------
score : double
The mean squared error between actual and predicted
"""
return np.mean(self.se(actual, predicted))
def rmse(self, actual, predicted):
"""
Computes the root mean squared error.
This function computes the root mean squared error between two lists
of numbers.
Parameters
----------
actual : list of numbers, numpy array
The ground truth value
predicted : same type as actual
The predicted value
Returns
-------
score : double
The root mean squared error between actual and predicted
"""
return np.sqrt(self.mse(actual, predicted))
def rmsle(self, actual, predicted):
"""
Computes the root mean squared log error.
This function computes the root mean squared log error between two lists
of numbers.
Parameters
----------
actual : list of numbers, numpy array
The ground truth value
predicted : same type as actual
The predicted value
Returns
-------
score : double
The root mean squared log error between actual and predicted
"""
return np.sqrt(self.msle(actual, predicted))
def model(self):
x = self.train[self.features].values
x_test = self.test[self.features].values
# TODO compute from no. of rows & features
k_folds = 10
n_times = 1
# dx = xgb.DMatrix(x, label=self.y, missing = float('nan'))
# dtest = xgb.DMatrix(x_test, missing = float('nan') )
# # setup parameters for xgboost
# param = {}
# param['objective'] = 'binary:logistic'
# param['eval_metric'] = 'error'
# # scale weight of positive examples
# param['eta'] = 0.3
# param['max_depth'] = 4
# param['silent'] = 1
## param['subsample'] = 0.5
## param['colsample_bytree'] = 0.6
scores = []
iters = []
for n in range(n_times):
# print '\n==============' + str(n) + '\n'
skf = StratifiedKFold(self.y, n_folds=k_folds, shuffle=True)
# for train_index, validation_index in skf:
for validation_index, train_index in skf:
x_train, x_validate = x[train_index], x[validation_index]
y_train, y_validate = self.y[train_index], self.y[validation_index]
# Initialize a classifier with key word arguments
clf = GradientBoostingRegressor()
clf.fit(x_train,y_train)
y_predicted = clf.predict(x_validate)
if self.eval_metric == 'rmse':
rmse_score = self.rmse(y_validate, y_predicted)
scores.append(rmse_score)
elif self.eval_metric == 'rmsle':
rmsle_score = self.rmsle(y_validate, y_predicted)
scores.append(rmsle_score)
# dtrain = xgb.DMatrix(x_train, label=y_train, missing = float('nan') )
# dval = xgb.DMatrix(x_validate, label=y_validate, missing = float('nan') ) #
#
# watchlist = [ (dtrain,'train'), (dval, 'test') ]
# num_round = 500
# clf = xgb.train(param, dtrain, num_round, watchlist, early_stopping_rounds=30, verbose_eval=False) #
# scores.append(np.absolute(clf.best_score))
# iters.append(clf.best_iteration)
print('features used : ' + ' '.join(self.features) )
print('overall ' + self.eval_metric + ' for ' + str(k_folds) + ' folds = ' + str(np.mean(scores)))
# print(param)
# print('XGBoost classifier = ' + str(1-np.mean(scores)) + ' +/- ' + str(round(np.std(scores)*100, 2)) + '%' )
# n_rounds_max = np.max(iters) + 1
#
## xgb_clf = xgb.train(param, dx, n_rounds_max)
# self.predictions = xgb_clf.predict(dtest)
#
# self.predictions[self.predictions >= 0.5] = 1
# self.predictions[self.predictions < 0.5] = 0
#
# submission = pd.DataFrame({ self.id_col: self.test['PassengerId'],
# self.y.name: self.predictions })
# submission[self.y.name] = submission[self.y.name].astype(int)
# submission.to_csv(os.path.join(self.directory, 'submission.csv' ), index=False)
class BinaryClassifier(Classifier):
def __init__(self, y, featureMixer, id_col):
self.directory = featureMixer.directory
self.id_col = id_col
self.y = y
self.train = featureMixer.train
self.test = featureMixer.test
self.features = featureMixer.features
def model(self):
x = self.train[self.features].values
x_test = self.test[self.features].values
# TODO compute from no. of rows & features
k_folds = 10
n_times = 1
dx = xgb.DMatrix(x, label=self.y, missing = float('nan'))
dtest = xgb.DMatrix(x_test, missing = float('nan') )
# setup parameters for xgboost
param = {}
param['objective'] = 'binary:logistic'
param['eval_metric'] = 'error'
# scale weight of positive examples
param['eta'] = 0.3
param['max_depth'] = 4
param['silent'] = 1
# param['subsample'] = 0.5
# param['colsample_bytree'] = 0.6
scores = []
iters = []
for n in range(n_times):
# print '\n==============' + str(n) + '\n'
skf = StratifiedKFold(self.y, n_folds=k_folds, shuffle=True)
# for train_index, validation_index in skf:
for validation_index, train_index in skf:
x_train, x_validate = x[train_index], x[validation_index]
y_train, y_validate = self.y[train_index], self.y[validation_index]
dtrain = xgb.DMatrix(x_train, label=y_train, missing = float('nan') )
dval = xgb.DMatrix(x_validate, label=y_validate, missing = float('nan') ) #
watchlist = [ (dtrain,'train'), (dval, 'test') ]
num_round = 500
clf = xgb.train(param, dtrain, num_round, watchlist, early_stopping_rounds=30, verbose_eval=False) #
scores.append(np.absolute(clf.best_score))
iters.append(clf.best_iteration)
print('\n=========== overall eval metric for ' + str(k_folds) + ' folds ===========')
print(' '.join(self.features) + '\n')
print(param)
print('XGBoost classifier = ' + str(1-np.mean(scores)) + ' +/- ' + str(round(np.std(scores)*100, 2)) + '%' )
n_rounds_max = np.max(iters) + 1
xgb_clf = xgb.train(param, dx, n_rounds_max)
self.predictions = xgb_clf.predict(dtest)
self.predictions[self.predictions >= 0.5] = 1
self.predictions[self.predictions < 0.5] = 0
submission = pd.DataFrame({ self.id_col: self.test[self.id_col],
self.y.name: self.predictions })
submission[self.y.name] = submission[self.y.name].astype(int)
submission.to_csv(os.path.join(self.directory, 'submission.csv' ), index=False)
class MulticlassClassifier(Classifier):
def __init__(self, y, featureMixer, id_col):
self.directory = featureMixer.directory
self.id_col = id_col
self.y = y
dtype = y.dtype.name
if dtype == 'object':
print y.name + ' is being encoded...'
le = LabelEncoder()
self.y = le.fit_transform(y.values)
self.noOfClasses = len( y.unique() )
self.train = featureMixer.train
self.test = featureMixer.test
self.features = featureMixer.features
def model(self):
x = self.train[self.features].values
x_test = self.test[self.features].values
# TODO compute from no. of rows & features
k_folds = 10
n_times = 1
dx = xgb.DMatrix(x, label=self.y, missing = float('nan'))
dtest = xgb.DMatrix(x_test, missing = float('nan') )
# setup parameters for xgboost
param = {}
param['objective'] = 'multi:softmax'
param['eval_metric'] = 'mlogloss'
param['num_class'] = self.noOfClasses
param['eta'] = 0.3
param['max_depth'] = 4
param['silent'] = 1
# param['subsample'] = 0.5
# param['colsample_bytree'] = 0.6
scores = []
iters = []
for n in range(n_times):
print '---------------- ' + str(n+1)
skf = StratifiedKFold(self.y, n_folds=k_folds, shuffle=True)
# for train_index, validation_index in skf:
for validation_index, train_index in skf:
x_train, x_validate = x[train_index], x[validation_index]
y_train, y_validate = self.y[train_index], self.y[validation_index]
dtrain = xgb.DMatrix(x_train, label=y_train, missing = float('nan') )
dval = xgb.DMatrix(x_validate, label=y_validate, missing = float('nan') ) #
watchlist = [ (dtrain,'train'), (dval, 'test') ]
num_round = 500
clf = xgb.train(param, dtrain, num_round, watchlist, early_stopping_rounds=30, verbose_eval=False) #
scores.append(np.absolute(clf.best_score))
iters.append(clf.best_iteration)
print('\n=========== overall eval metric for ' + str(k_folds) + ' folds ===========')
print(' '.join(self.features) + '\n')
print(param)
print('XGBoost classifier = ' + str(1-np.mean(scores)) + ' +/- ' + str(round(np.std(scores)*100, 2)) + '%' )
n_rounds_max = np.max(iters) + 1
xgb_clf = xgb.train(param, dx, n_rounds_max)
self.predictions = xgb_clf.predict(dtest)
print self.predictions
submission = pd.DataFrame({ self.id_col: self.test[self.id_col],
self.y.name: self.predictions })
submission[self.y.name] = submission[self.y.name]
submission.to_csv(os.path.join(self.directory, 'submission.csv' ), index=False)
class Predictor(object):
def __init__(self, featureMixer):
self.predictors = {}
self.targets = featureMixer.targets
self.train = featureMixer.train
self.test = featureMixer.test
self.categoricals = featureMixer.categoricals
self.numericals = featureMixer.numericals
for target in self.targets:
y = self.train[target]
if target in self.categoricals:
self.predictors[y.name] = Classifier(y)
# self.__classification__(y)
else:
# self.__regression__()
self.predictors[y.name] = Regressor()
def computeFolds(self, nrows, nfeats):
pass
def __regression__(self):
# TODO a regression
print 'YO!!! regression'
def modeling(self):
for target, modeler in self.predictors:
print 'modeling ' + target
modeler.model()