forked from Aiwiscal/ECG-ML-DL-Algorithm-Python
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathClassification-DL_CNN.py
134 lines (112 loc) · 4.78 KB
/
Classification-DL_CNN.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
# -*- coding: utf-8 -*-
"""
Created on Thu May 17 21:01:53 2018
@author: Winham
===================基于1维CNN的ECG分类算法========================
*需要第三方工具包numpy,h5py,scikit-learn
*基于深度学习框架TensorFlow
*涉及函数的使用方法可自行查看工具包文档，baidu即可得
*ECG算法入门系列博客：https://blog.csdn.net/qq_15746879
*开源github：https://github.com/Aiwiscal
*本代码所需要的数据和标签文件来自matlab提取
*详情：https://blog.csdn.net/qq_15746879/article/details/80340958
==================================================================
"""
#载入所需工具包
import time
import numpy as np
import h5py as hp
import tensorflow as tf
from sklearn.metrics import confusion_matrix
sess=tf.InteractiveSession()
#载入.mat文件的函数,h5py解码并转换为numpy数组
def load_mat(path_data,name_data,dtype='float32'):
data=hp.File(path_data)
arrays_d={}
for k,v in data.items():
arrays_d[k]=np.array(v)
dataArr=np.array(arrays_d[name_data],dtype=dtype)
return dataArr
#使用TensorFlow组件完成CNN网络的搭建，与教程中参数略有不同
def CNNnet(inputs,n_class):
conv1 = tf.layers.conv1d(inputs=inputs, filters=4, kernel_size=31, strides=1, \
padding='same', activation = tf.nn.relu)
avg_pool_1 = tf.layers.average_pooling1d(inputs=conv1, pool_size=5, strides=5, \
padding='same')
conv2 = tf.layers.conv1d(inputs=avg_pool_1, filters=8, kernel_size=6, strides=1,\
padding='same', activation = tf.nn.relu)
avg_pool_2 = tf.layers.average_pooling1d(inputs=conv2, pool_size=5, strides=5,\
padding='same')
flat = tf.reshape(avg_pool_2, (-1, int(250/5/5*8)))
logits=tf.layers.dense(inputs=flat, units=n_class, activation=None)
logits=tf.nn.softmax(logits)
return logits
#随机获取一个batch大小的数据，用于训练
def get_batch(train_x,train_y,batch_size):
indices=np.random.choice(train_x.shape[0],batch_size,False)
batch_x=train_x[indices]
batch_y=train_y[indices]
return batch_x,batch_y
#设定路径及文件名并载入，这里的心拍在Matlab下截取完成
#详情：https://blog.csdn.net/qq_15746879/article/details/80340671
Path='F:/Python files/ECGPrimer/' #自定义路径要正确
DataFile='Data_CNN.mat'
LabelFile='Label_OneHot.mat'
print("Loading data and labels...")
tic=time.time()
Data=load_mat(Path+DataFile,'Data')
Label=load_mat(Path+LabelFile,'Label')
Data=Data.T
Indices=np.arange(Data.shape[0]) #随机打乱索引并切分训练集与测试集
np.random.shuffle(Indices)
print("Divide training and testing set...")
train_x=Data[Indices[:10000]]
train_y=Label[Indices[:10000]]
test_x=Data[Indices[10000:]]
test_y=Label[Indices[10000:]]
toc=time.time()
print("Elapsed time is %f sec."%(toc-tic))
print("======================================")
print("1D-CNN setup and initialize...")
tic=time.time()
x=tf.placeholder(tf.float32, [None, 250]) #定义placeholder数据入口
x_=tf.reshape(x,[-1,250,1])
y_=tf.placeholder(tf.float32,[None,4])
logits=CNNnet(x_,4)
learning_rate=0.01
batch_size=16
maxiters=15000
cost=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits,labels=y_))
#这里使用了自适应学习率的Adam训练方法，可以认为是SGD的高级演化版本之一
train_step=tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)
tf.global_variables_initializer().run()
toc=time.time()
print("Elapsed time is %f sec."%(toc-tic))
print("======================================")
print("1D-CNN training and testing...")
tic=time.time()
for i in range(maxiters):
batch_x,batch_y=get_batch(train_x,train_y,batch_size)
train_step.run(feed_dict={x:batch_x,y_:batch_y})
if i%500==0:
loss=cost.eval(feed_dict={x:train_x,y_:train_y})
print("Iteration %d/%d:loss %f"%(i,maxiters,loss))
y_pred=logits.eval(feed_dict={x:test_x,y_:test_y})
y_pred=np.argmax(y_pred,axis=1)
y_true=np.argmax(test_y,axis=1)
toc=time.time()
print("Elapsed time is %f sec."%(toc-tic))
Acc=np.mean(y_pred==y_true)
Conf_Mat=confusion_matrix(y_true,y_pred) #利用专用函数得到混淆矩阵
Acc_N=Conf_Mat[0][0]/np.sum(Conf_Mat[0])
Acc_V=Conf_Mat[1][1]/np.sum(Conf_Mat[1])
Acc_R=Conf_Mat[2][2]/np.sum(Conf_Mat[2])
Acc_L=Conf_Mat[3][3]/np.sum(Conf_Mat[3])
print('\nAccuracy=%.2f%%'%(Acc*100))
print('Accuracy_N=%.2f%%'%(Acc_N*100))
print('Accuracy_V=%.2f%%'%(Acc_V*100))
print('Accuracy_R=%.2f%%'%(Acc_R*100))
print('Accuracy_L=%.2f%%'%(Acc_L*100))
print('\nConfusion Matrix:\n')
print(Conf_Mat)
print("======================================")