-
Notifications
You must be signed in to change notification settings - Fork 0
/
SSI_1DCNN-LSTM.py
256 lines (192 loc) · 7.57 KB
/
SSI_1DCNN-LSTM.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
# -*- coding: utf-8 -*-
"""
Created on Tue Jun 8 15:58:43 2021
@author: michi
"""
# -*- coding: utf-8 -*-
"""
Created on Tue Jun 8 14:32:55 2021
@author: michi
"""
# cnn model
import matplotlib.pyplot as plt
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Flatten
#from keras.layers import Dropout
from keras.layers.convolutional import Conv1D
from keras.layers.convolutional import MaxPooling1D
from keras.utils import to_categorical
from keras.layers import LSTM
from keras.layers import TimeDistributed
from sklearn.model_selection import train_test_split
import os
import scipy.io as spio
import numpy as np
from keras.callbacks import EarlyStopping
from keras.layers import Dropout
#from sklearn.metrics import confusion_matrix
from sklearn.metrics import precision_recall_fscore_support
#from sklearn.metrics import confusion_matrix
from sklearn.model_selection import (TimeSeriesSplit, KFold, ShuffleSplit,
StratifiedKFold, GroupShuffleSplit,
GroupKFold, StratifiedShuffleSplit)
## Settings
seconds = 3
expectedSampleRate = 100
epochLength = seconds*expectedSampleRate
inputTimeSerie = 'markervel'
inputColumns = [3,4,5,42,43,44]
n_features = len(inputColumns)
#inputTimeSerie = 'markerpos'
############# Loading the data ###############################
filename = []
data = []
#dataDir = 'D:\\Stroke_data\\V2_Data\\10sec\\ML_markervel\\ParYpsi\\'
dataDir = 'F:\\SSI_data\\10sec\\ML_markervel\\ParYpsi\\'
#file = 'S001RQ1_MLPert1_Trial1_ParYpsi_FR0_V1'
group = []
for file in os.listdir(dataDir):
data.append(spio.loadmat(dataDir + file))
filename.append(file)
group.append(int(file[1:4]))
group = np.array(group)
#### first attempt.
# we are going for the RANK. --? columns 43/45
# get the dict out of the list
for numtrials in range(len(filename)):
data[numtrials] = (data[numtrials][inputTimeSerie])
##### For different inputs, different handling of the data.
#### Stack the data of multiple trials on top of each other and perform normalisation.
if inputTimeSerie == 'markerpos':
tempData = np.zeros(epochLength)
tempData1 = []
X_data = [0,0,0] #np.zeros((1001*177,3))
Norx = [0,0,0]
for indx in range(len(data)):
for j in range(0,3):
x = data[indx][0:epochLength,j+42]
test = (x-min(x))/(max(x)-min(x))
tempData = np.vstack((tempData,test))
tempData = np.delete(tempData, (0), axis=0)
tempData = np.transpose(tempData)
X_data = np.vstack((X_data,tempData))
tempData = np.zeros(epochLength)
if inputTimeSerie == 'markervel':
X_data = np.zeros(len(inputColumns))#[0,0,0]
for indx in range(len(data)):
x = data[indx][0:epochLength,inputColumns]
# x = data[indx]['markervel'][0:epochLength,42:45]
X_data = np.vstack((X_data,x))
### Always remove first row (due to initialize issues)
X_data = np.delete(X_data, (0), axis=0)
##### Get the dependent y data from filenames. #######
y = [0]
for indx in range(len(filename)):
x = filename[indx].split('FR')
y = np.vstack((y,int(x[1][0])))
y = np.delete(y,(0),axis=0)
###### Make y a categorical value #####
y_cat = to_categorical(y)
######## Reshape the X data ############
X_data = X_data.reshape(len(y_cat),epochLength,n_features)
####### Split data by group (subject level) #################
group_kfold = GroupKFold(n_splits=2)
group_kfold.get_n_splits(X_data, y_cat, group)
gss = GroupShuffleSplit(n_splits=2, train_size=.7, random_state=42)
for train_index, test_index in gss.split(X_data, y_cat, group):
# print("TRAIN:", train_index, "TEST:", test_index)
X_train, X_test = X_data[train_index], X_data[test_index]
y_train, y_test = y_cat[train_index], y_cat[test_index]
# print(X_train, X_test, y_train, y_test)
####### Split data by trial (within subject level) #################
#X_train, X_test, y_train, y_test = train_test_split(X_data, y_cat, test_size=0.3, random_state=42)
###################################################################################
# reshape data into time steps of sub-sequences for the lstm
n_steps, n_length = seconds, expectedSampleRate
X_train = X_train.reshape((X_train.shape[0], n_steps, n_length, n_features))
X_test = X_test.reshape((X_test.shape[0], n_steps, n_length, n_features))
n_outputs = y_train.shape[1]
## Determine imbalance and determine weights
n_samples = len(y)
n_classes = 2
n_samplesj = np.sum(y == 0)
n_samplesi = np.sum(y == 1)
#n_samplesk = np.sum(finalDatay == 2)
#n_samplesg = np.sum(finalDatay == 3)
wj=n_samples / (n_classes * n_samplesj)
wi=n_samples / (n_classes * n_samplesi)
#wk=n_samples / (n_classes * n_samplesk)
#wg=n_samples / (n_classes * n_samplesg)
class_weight = {0: wj,
1: wi
}
## Define the model.
model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=3, activation='relu'), input_shape=(None,n_length,n_features)))
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=3, activation='relu')))
model.add(TimeDistributed(Dropout(0.5)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dropout(0.5))
model.add(Dense(100, activation='relu'))
model.add(Dense(n_outputs, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
print(model.summary())
## Run the training part
initial_weights = model.get_weights()
train_sizes = (len(X_train) * np.linspace(0.1,0.999,4)).astype(int)
train_scores = []
test_scores = []
for train_size in train_sizes:
X_train_frac, _, y_train_frac, _ = train_test_split(X_train,y_train, train_size=train_size)
# at each iteration, set original weights
# to the initial random weights
print('check1')
model.set_weights(initial_weights)
print('check2')
h = model.fit(X_train_frac, y_train_frac,
verbose=1,
epochs=100,
callbacks=[EarlyStopping(monitor='loss', patience=5)],class_weight=class_weight)#,
#class_weight=class_weights
print('check3')
r = model.evaluate(X_train_frac, y_train_frac, verbose = 0)
print('check4')
train_scores.append(r[-1])
print('check5')
e = model.evaluate(X_test, y_test, verbose=0)
test_scores.append(e[-1])
print ("Done size: ", train_size )
plt.figure()
plt.plot(train_sizes, train_scores, 'o-', label="training score")
plt.plot(train_sizes, test_scores, 'o-',label="Cross validation scores")
plt.legend(loc="best")
print(model.evaluate(X_test,y_test))
predicted_labels = model.predict(X_test)
predicted_labels = np.argmax(predicted_labels, axis=1)
precision_recall_fscore_support
y_train_count = np.argmax(y_train, axis = 1)
unique, counts = np.unique(y_train_count, return_counts=True)
dict(zip(unique, counts)) # MAKE A BAR PLOT HERE TO PUT INTO THE PRESENTATION.
y_pos = np.arange(len(counts))
plt.figure(0, figsize=(15, 8))
#fig = plt.figure(0)
plt.bar(y_pos, counts, align='center', alpha=0.5)
plt.xticks(y_pos, unique)
plt.ylabel('Number of Observations')
plt.title('Class observations')
plt.show()
#y_test = np.argmax(y_test, axis=1)
#cm = confusion_matrix(y_test, predicted_labels)
#print('Confusion matrix')
#fig2, ax = plt.subplots()
#plot_confusion_matrix(cm)
##fig2.savefig(Name2)
#cm1 = cm / cm.astype(np.float).sum(axis=1)
#print('Confusion matrix, with normalization')
#fig, ax = plt.subplots()
#plot_confusion_matrix(cm1)
#plt.show()
#performance = precision_recall_fscore_support(y_test, predicted_labels, average='macro')