-
Notifications
You must be signed in to change notification settings - Fork 4
/
get_all_features.py
189 lines (170 loc) · 7.95 KB
/
get_all_features.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
import scipy.io as sio
import numpy as np
import glob
import os
import scipy.misc as misc
from sklearn import preprocessing
import scipy as sp
import scipy.signal as spsig
import pandas as pd
import scipy.stats as spstat
import json
#def get_data(file):
# matfile = sio.loadmat(file)
# data = (matfile['dataStruct']['data'][0,0]).T
# return data
def get_data(file):
try:
matfile = sio.loadmat(file)
data = (matfile['dataStruct']['data'][0,0]).T
return data
except Exception:
print 'bad file:', file
return np.zeros([16,400*10*60])
def long_features(pat, outfile, datapath):
#pat = 3
#outfile='pat_'+str(pat)+'_long_newtest_sub.csv'
# file path for the new_test data
#f = '/mnt/am02_scratch/blang/kaggle_data/test_'+str(pat)+'_new/*mat'
# file path for the training and hold-out testing
#f = '/mnt/am02_scratch/blang/kaggle_data/CV/pat_'+str(pat)+'/train/*mat'
f = datapath + '/*mat'
pat_num = pat
ff = glob.glob(f)
label=[str(os.path.basename(n)) for n in ff]
output = []
featureList = []
mydata = []
bands=[0.1,4,8,12,30,70]
for i in range(len(ff)):
print float(i)/float(len(ff))
output = []
featureList = []
if os.path.basename(ff[i]) == '1_45_1.mat':
continue
data = get_data(ff[i])
data = preprocessing.scale(data, axis=1,with_std=True)
featureList.append('File')
output.append(label[i])
featureList.append('pat')
output.append(pat_num)
# get correlation Coef. this will be 16x16
h=np.corrcoef(data)
h=np.nan_to_num(h)
# only want upper triangle
ind = np.triu_indices(16, 1)
htri = h[ind]
for ii in range(np.size(htri)):
featureList.append('coef%i'%(ii))
output.append(htri[ii])
c,v = np.linalg.eig(h)
c.sort()
c = np.real(c)
for e in range(len(c)):
featureList.append('coef_timeEig%i'%(e))
output.append(c[e])
for j in range(16):
hold = spsig.decimate(data[j,:],5,zero_phase=True)
featureList.append('sigma%i'%(j))
output.append(hold.std())
featureList.append('kurt%i'%(j))
output.append(spstat.kurtosis(hold))
featureList.append('skew%i'%(j))
output.append(spstat.skew(hold))
featureList.append('zero%i'%(j))
output.append(((hold[:-1] * hold[1:]) < 0).sum())
diff = np.diff(hold, n=1)
diff2 = np.diff(hold,n=2)
featureList.append('sigmad1%i'%(j))
output.append(diff.std())
featureList.append('sigmad2%i'%(j))
output.append(diff2.std())
featureList.append('zerod%i'%(j))
output.append(((diff[:-1] * diff[1:]) < 0).sum())
featureList.append('zerod2%i'%(j))
output.append(((diff2[:-1] * diff2[1:]) < 0).sum())
featureList.append('RMS%i'%(j))
output.append(np.sqrt((hold**2).mean()))
f, psd = spsig.welch(hold, fs = 80)
psd[0] = 0
featureList.append('MaxF%i'%(j))
output.append(psd.argmax())
featureList.append('SumEnergy%i'%(j))
output.append(psd.sum())
psd /= psd.sum()
for c in range(1,len(bands)):
featureList.append('BandEnergy%i%i'%(j,c))
output.append(psd[(f>bands[c-1])&(f<bands[c])].sum())
featureList.append('entropy%i'%(j))
output.append(-1.0*np.sum(psd[f>bands[0]]*np.log10(psd[f>bands[0]])))
#pdb.exit()
featureList.append('Mobility%i'%(j))
output.append(np.std(diff)/hold.std())
featureList.append('Complexity%i'%(j))
output.append(np.std(diff2)*np.std(hold)/(np.std(diff)**2.))
mydata.append(pd.DataFrame({'Features':output},index=featureList).T)
trainSample = pd.concat(mydata,ignore_index=True)
trainSample.to_csv(outfile)
return 1
def short_features(pat, outfile, datapath):
#pat = 3
#outfile = 'pat_'+str(pat)+'_short_newtest_sub.csv'
#file path for training and hold-out testing
#f = '/mnt/am02_scratch/blang/kaggle_data/CV/pat_'+str(pat)+'/train/*mat'
#file path for the new testing
#f = '/mnt/am02_scratch/blang/kaggle_data/test_'+str(pat)+'_new/*mat'
f = datapath+'/*mat'
ff = glob.glob(f)
label=[str(os.path.basename(n)) for n in ff]
output = []
featureList = []
mydata = []
bands=[0.1,4,8,12,30,70,180]
rate = 400.
for i in range(len(ff)):
print float(i)/float(len(ff))
data_full = get_data(ff[i])
output = []
featureList = []
featureList.append('File')
output.append(label[i])
featureList.append('pat')
output.append(pat)
for j in range(19):
if os.path.basename(ff[i]) == '1_45_1.mat':
continue
data = data_full[:,j*int(rate*60/2):(j)*int(rate*60/2) + int(rate)*60]
data = preprocessing.scale(data, axis=1,with_std=True)
for k in range(16):
hold = data[k,:]
f,psd = spsig.welch(hold, fs=400, nperseg=2000)
psd = np.nan_to_num(psd)
psd /= psd.sum()
for c in range(1,len(bands)):
featureList.append('BandEnergy_%i_%i_%i'%(j,k,c))
output.append(psd[(f>bands[c-1])&(f<bands[c])].sum())
mydata.append(pd.DataFrame({'Features':output},index=featureList).T)
trainSample = pd.concat(mydata,ignore_index=True)
trainSample.to_csv(outfile)
return 1
def main():
feat = json.load(open('SETTINGS.json'))
keys = feat.keys()
pat = feat['pat']
if feat['make_test'] == 1:
outfile = feat['feat']+'/pat_'+str(pat)+'_long_newtest_sub.csv'
l = long_features(pat,outfile,feat['test'])
outfile = feat['feat']+'/pat_'+str(pat)+'_short_newtest_sub.csv'
s = short_features(pat, outfile, feat['test'])
if feat['make_train'] == 1:
outfile = feat['feat']+'/pat_'+str(pat)+'_long_train.csv'
l = long_features(pat,outfile,feat['train'])
outfile = feat['feat'] + '/pat_'+str(pat)+'_short_train.csv'
s = short_features(pat, outfile, feat['train'])
if feat['make_hold'] == 1:
outfile = feat['feat']+'/pat_'+str(pat)+'_long_test.csv'
l = long_features(pat,outfile,feat['hold-out'])
outfile = feat['feat'] + '/pat_'+str(pat)+'_short_test.csv'
s = short_features(pat, outfile, feat['hold-out'])
if __name__ == "__main__":
main()