-
Notifications
You must be signed in to change notification settings - Fork 0
/
voting_ensemble.py
224 lines (163 loc) · 6.46 KB
/
voting_ensemble.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
from sklearn.base import BaseEstimator
from sklearn.base import ClassifierMixin
from sklearn.metrics import classification_report, confusion_matrix, accuracy_score
import numpy as np
import operator
import glob
import warnings
def get_predictions(model, X):
if hasattr(model, 'predict_proba'): # Normal SKLearn classifiers
pred = model.predict_proba(X)
elif hasattr(model, '_predict_proba_lr'): # SVMs
pred = model._predict_proba_lr(X)
else:
pred = model.predict(X)
if len(pred.shape) == 1: # for 1-d ouputs
pred = pred[:, None]
return pred
def check_module_exists(modulename):
try:
__import__(modulename)
except ImportError:
return False
return True
if check_module_exists('xgboost'):
import xgboost as xgb
class SoftVoteClassifier(BaseEstimator, ClassifierMixin):
"""
Ensemble classifier for pre-trained scikit-learn estimators.
Parameters
----------
clf : tuple - (classifier_name, clf)
A list of pre-trained scikit-learn classifier objects.
Can include XGBoost models at the very end of the classifier list.
weights : `list` (default: `None`)
If `None`, the majority rule voting will be applied to the predicted class labels.
If a list of weights (`float` or `int`) is provided, the averaged raw probabilities (via `predict_proba`)
will be used to determine the most confident class label.
normalize_weights : bool (default: False)
If True, will normalize the weights provided so that they sum up to 1.0
verbose : bool (default: False)
If True, will print out information about model prediction
"""
def __init__(self, clfs, weights=None, normalize_weights=False, verbose=False):
self.clfs = clfs
self.verbose = verbose
self.normalize_weights = normalize_weights
if self.normalize_weights:
weight_sum = np.sum(np.asarray(weights))
weights = [w / weight_sum for w in weights]
self.weights = weights
else:
self.weights = weights
def fit(self, X, y):
"""
Fit the scikit-learn estimators.
Parameters
----------
X : numpy array, shape = [n_samples, n_features]
Training data
y : list or numpy array, shape = [n_samples]
Class labels
"""
warnings.warn('SoftVoteClassifier simply votes on pre-trained models. Please provide pre-trained model '
'in the classifier list.')
def predict(self, X):
"""
Parameters
----------
X : numpy array, shape = [n_samples, n_features]
Returns
----------
maj : list or numpy array, shape = [n_samples]
Predicted class labels by majority rule
"""
self.probas_ = []
data = X
for name, clf in self.clfs:
if check_module_exists('xgboost'):
if isinstance(clf, xgb.Booster): # XGBoost model, needs DMatrix for data
if not hasattr(data, 'feature_names'): # data is numpy array, convert to DMatrix
data = xgb.DMatrix(X)
self.probas_.append(get_predictions(clf, data))
if self.verbose:
print('Obtained predictions of model %s' % name)
avg = np.average(self.probas_, axis=0, weights=self.weights)
# self.class_probas_ = np.asarray([get_predictions(clf, X) for clf in self.clfs])
# if self.weights:
# #maj = np.apply_along_axis(lambda x: max(enumerate(x), key=operator.itemgetter(1))[0], axis=1, arr=self.class_probas_)
# avg = np.average(self.probas_, axis=0, weights=self.weights)
# else:
# #maj = np.asarray([np.argmax(np.bincount(self.class_probas_[:, c])) for c in range(self.class_probas_.shape[1])])
#
maj = np.argmax(avg, axis=1)
return maj
def predict_proba(self, X):
"""
Parameters
----------
X : numpy array, shape = [n_samples, n_features]
Returns
----------
avg : list or numpy array, shape = [n_samples, n_probabilities]
Weighted average probability for each class per sample.
"""
self.probas_ = []
data = X
for name, clf in self.clfs:
if check_module_exists('xgboost'):
if isinstance(clf, xgb.Booster): # XGBoost model, needs DMatrix for data
if not hasattr(data, 'feature_names'): # data is numpy array, convert to DMatrix
data = xgb.DMatrix(X)
self.probas_.append(get_predictions(clf, data))
if self.verbose:
print('Obtained predictions of model %s' % name)
avg = np.average(self.probas_, axis=0, weights=self.weights)
return avg
def predict_dir(self, dir):
"""
Parameters
----------
dir : a directory containing the numpy arrays which contain the predictions
Returns
----------
maj : list or numpy array, shape = [n_samples]
Predicted class labels by majority rule
"""
self.probas_ = []
path = dir + "*.npy"
files = glob.glob(path)
for i, fn in enumerate(files):
if 'voting' in fn:
continue
preds = np.load(fn)
preds = preds.mean(axis=0)
self.probas_.append(preds)
if self.verbose:
print('Obtained predictions of model %d' % (i + 1))
avg = np.average(self.probas_, axis=0, weights=self.weights)
maj = np.argmax(avg, axis=1)
return maj
def predict_proba_dir(self, dir):
"""
Parameters
----------
dir : a directory containing the numpy arrays which contain the predictions
Returns
----------
avg : list or numpy array, shape = [n_samples, n_probabilities]
Weighted average probability for each class per sample.
"""
self.probas_ = []
path = dir + "*.npy"
files = glob.glob(path)
for i, fn in enumerate(files):
if 'voting' in fn:
continue
preds = np.load(fn)
preds = preds.mean(axis=0)
self.probas_.append(preds)
if self.verbose:
print('Obtained predictions of model %d' % (i + 1))
avg = np.average(self.probas_, axis=0, weights=self.weights)
return avg