/
SVM_Results.py
185 lines (132 loc) · 6.96 KB
/
SVM_Results.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
__author__ = 'xie'
## This file load the picked data, and perform random split,
# to form training, test, and validation set
import cPickle
import numpy as np
from random import randrange
import matplotlib.pyplot as plt
from sklearn.pipeline import Pipeline
from sklearn import preprocessing
from sklearn import svm
from sklearn import cross_validation
from sklearn.cross_validation import train_test_split
from sklearn.learning_curve import learning_curve
from sklearn.neural_network import BernoulliRBM
def get_train_test(data):
features = data[0]
labels = data[1]
#flatten feature to 2d matrix
flat_features = features.reshape(384, 60*40*3)
seed = randrange(100)
train_x, test_x, train_y, test_y = train_test_split(flat_features, labels,
test_size=0.2,
random_state=seed)
return train_x, test_x, train_y, test_y
def simpleSVM(trainfeatures, testfeatures, trainlabels, testlabels):
## ******************* Feature Scaling *******************
#print "performing feature scaling"
min_max_scaler = preprocessing.MinMaxScaler()
trainfeatures_fs = min_max_scaler.fit_transform(trainfeatures)
testfeatures_fs = min_max_scaler.transform(testfeatures)
# Training
#print "training SVM model"
clf = svm.SVC(C=5.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=10.0,
shrinking=True, probability=False, tol=0.001, cache_size=200,
class_weight=None, verbose=False, max_iter=-1, random_state=None)
# clf = svm.SVC(C=5.0, kernel='rbf', degree=3, gamma=2, coef0=10, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
clf.fit(trainfeatures_fs, trainlabels)
results = clf.predict(testfeatures_fs)
results = results.ravel()
testerror = float(len(testlabels)
- np.sum(testlabels == results))/float(len(testlabels))
# print"error rate with SVM 2 is %.4f" %testerror
return testerror
def RBM_SVM(trainfeatures, testfeatures, trainlabels, testlabels):
# ******************* Scikit-learning RBM + SVM *******************
print "train RBM+SVM model"
## trainfeatures = (trainfeatures - np.min(trainfeatures, 0)) / (np.max(trainfeatures, 0) + 0.0001) # 0-1 scaling
min_max_scaler = preprocessing.MinMaxScaler()
trainfeatures_fs = min_max_scaler.fit_transform(trainfeatures)
testfeatures_fs = min_max_scaler.transform(testfeatures)
# SVM parameters
clf = svm.SVC(C=5.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=10.0,
shrinking=True, probability=False, tol=0.001, cache_size=200,
class_weight=None, verbose=False, max_iter=-1, random_state=None)
# RBM parameters
rbm = BernoulliRBM(random_state=0, verbose=True)
rbm.learning_rate = 0.06
rbm.n_iter = 20
# Machine learning pipeline
classifier = Pipeline(steps=[('rbm', rbm), ('svm', clf)])
# More components tend to give better prediction performance, but larger
# fitting time
rbm.n_components = 400
classifier.fit(trainfeatures_fs, trainlabels)
results = classifier.predict(testfeatures_fs)
results = results.ravel()
testerror = float(len(testlabels)
- np.sum(testlabels == results))/float(len(testlabels))
# print"error rate with SVM is %.4f" %testerror
return testerror
if __name__ == '__main__':
# folder = os.path.dirname(__file__)
folder = "c:/users/xie/playground/cctv classification"
pickle_file = folder+"/pickle_data/data.pkl"
load_file = open(pickle_file, 'rb')
data = cPickle.load(load_file)
labels = data[1]
features = data[0]
rates = []
for i in xrange(1, 20):
trainfeatures, testfeatures, trainlabels, testlabels = get_train_test(data)
# testerror=simpleSVM(trainfeatures, testfeatures, trainlabels, testlabels)
testerror = RBM_SVM(trainfeatures, testfeatures, trainlabels, testlabels)
print "test run %d, error rate is %1.4f " % (i, testerror)
rates.append(testerror)
mean_error = np.array(rates).mean()
print "average error rate is %1.4f" % mean_error
# trainfeatures, testfeatures, trainlabels, testlabels=get_train_test(data)
# test run 1: average error rate: 0.4252
# clf = svm.SVC(C=5.0, kernel='poly', degree=3, gamma=0.5, coef0=10.0, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 2: average error rate: 0.4087
# clf = svm.SVC(C=5.0, kernel='rbf', degree=3, gamma=0.5, coef0=10.0, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 3: average error rate: 0.3841, 0.3855 0.4224 0.3978 0.4040 0.4163 0.3835 ----- best sigmoid so far
# clf = svm.SVC(C=5.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=10.0, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 4: average error rate: 0.4087
# clf = svm.SVC(C=5.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=15.0, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 5: average error rate: 0.4033
# clf = svm.SVC(C=5.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=7.5, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 6: average error rate: 0.4108
# clf = svm.SVC(C=3.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=10, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 7: average error rate: 0.4156
# clf = svm.SVC(C=7.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=10, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# test run 8: average error rate: 0.3923, 0.4060
# clf = svm.SVC(C=5.0, kernel='rbf', degree=3, gamma=2, coef0=10, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
# RBM_SVM test_run 1: 0.4053, 0.4060
# # SVM parameters
# clf = svm.SVC(C=5.0, kernel='sigmoid', degree=3, gamma=0.5, coef0=10.0, shrinking=True,
# probability=True, tol=0.001, cache_size=200, class_weight=None, verbose=False,
# max_iter=-1, random_state=None)
#
# # RBM parameters
# rbm = BernoulliRBM(random_state=0, verbose=True)
# rbm.learning_rate = 0.06
# rbm.n_iter = 20