forked from dydx-git/COVID-Classifier
-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_model.py
139 lines (124 loc) · 6.21 KB
/
train_model.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
from utils import *
#from extract_features import extractFeaturesTest
# =============================================================================
# source_dir
# =============================================================================
source_dir='./'
# =============================================================================
# load mat files
# =============================================================================
covid_features=sio.loadmat(os.path.join(source_dir,'covid.mat'))
covid_features=covid_features['covid']
normal_features=sio.loadmat(os.path.join(source_dir,'normal.mat'))
normal_features=normal_features['normal']
pneumonia_features=sio.loadmat(os.path.join(source_dir,'pneumonia.mat'))
pneumonia_features=pneumonia_features['pneumonia']
# =============================================================================
# devide feature pool into inputs and target labels
# =============================================================================
X=np.concatenate((covid_features[:,:-1],normal_features[:,:-1],pneumonia_features[:,:-1]), axis=0)#inputs
y=np.concatenate((covid_features[:,-1],normal_features[:,-1],pneumonia_features[:,-1]), axis=0)#target labels
# =============================================================================
# normalization
# =============================================================================
min_max_scaler=MinMaxScaler()
X = min_max_scaler.fit_transform(X)
# =============================================================================
# feature reduction (K-PCA)
# =============================================================================
transformer = KernelPCA(n_components=64, kernel='linear')
X = transformer.fit_transform(X)
print("X shape after KernelPCA: ",X.shape)
# =============================================================================
# devide data into test,train, and validation sets
# =============================================================================
y = to_categorical(y)
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.2, random_state=1)
X_train, X_val, y_train, y_val = train_test_split(
X_train, y_train, test_size=0.25, random_state=1)
# =============================================================================
# build model
# =============================================================================
model = build_model(feature_size=X.shape[-1], n_classes=y.shape[-1])
show_model(model)
# =============================================================================
# train model
# =============================================================================
opt = tf.keras.optimizers.Adam(lr=0.001)
criterion = tf.keras.losses.categorical_crossentropy
model.compile(optimizer=opt, loss=criterion,metrics=[keras.metrics.categorical_accuracy,keras.metrics.Recall(name="recall"),keras.metrics.Precision(name="precision")])
model_early_stopping=EarlyStopping(monitor='val_loss', min_delta=.005, patience=10, verbose=1)# early stopping settings
model.fit(X_train, y_train,batch_size = 2,epochs=100, validation_data = (X_val, y_val),callbacks=[model_early_stopping])
model.save('saved_model/my_model')
fig = plt.figure()
plt.plot(model.history.history['val_loss'], 'r', model.history.history['loss'], 'b')
plt.xlabel('Epochs')
plt.ylabel('Loss Score')
plt.grid(1)
plt.savefig('training_loss.jpg',dpi=300)
# =============================================================================
# evaluate model
# =============================================================================
Y_Score=model.predict(X_test)
print("X_test: ",X_test)
print("Y_score: ",Y_Score)
y_pred = np.argmax(Y_Score, axis=1)
cm=confusion_matrix(np.argmax(y_test, axis=1),y_pred)
print(cm)
fig = plt.figure()
plot_confusion_matrix(cm,classes=['COVID-19','Normal','Pneumonia'])
plt.savefig('conf_matrix.jpg',dpi=300)
test_loss=model.evaluate(X_test,y_test,verbose=1)#evaluate model
print(test_loss)#print test loss and metrics information
# =============================================================================
# ROC curve where positive label is COVID-19
# =============================================================================
pos_label=0
fpr, tpr, thresholds = roc_curve(np.argmax(y_test, axis=1), Y_Score[:,pos_label], pos_label=pos_label)
roc_auc = auc(fpr, tpr)# calculate auc value
fig = plt.figure()
plt.plot(fpr, tpr, color='darkorange', label='ROC curve (area = %0.2f)\npos_label=COVID-19' % roc_auc)
plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver operating characteristic')
plt.grid(1)
plt.legend(loc="lower right")
plt.savefig('roc_covid.jpg',dpi=300)
# =============================================================================
# ROC curve where positive label is normal
# =============================================================================
pos_label=1
fpr, tpr, thresholds = roc_curve(np.argmax(y_test, axis=1), Y_Score[:,pos_label], pos_label=pos_label)
roc_auc = auc(fpr, tpr)# calculate auc value
fig = plt.figure()
plt.plot(fpr, tpr, color='darkorange', label='ROC curve (area = %0.2f)\npos_label=Normal' % roc_auc)
plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver operating characteristic')
plt.grid(1)
plt.legend(loc="lower right")
plt.savefig('roc_normal.jpg',dpi=300)
# =============================================================================
# ROC curve where positive label is COVID-19
# =============================================================================
pos_label=2
fpr, tpr, thresholds = roc_curve(np.argmax(y_test, axis=1), Y_Score[:,pos_label], pos_label=pos_label)
roc_auc = auc(fpr, tpr)# calculate auc value
fig = plt.figure()
plt.plot(fpr, tpr, color='darkorange', label='ROC curve (area = %0.2f)\npos_label=Pneumonia' % roc_auc)
plt.plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver operating characteristic')
plt.grid(1)
plt.legend(loc="lower right")
plt.savefig('roc_pneumonia.jpg',dpi=300)