简单神经网络训练模型(pima-indians-diabetes)

数据集

下载&查看
Pima Indians Diabetes Data Set(皮马印第安人糖尿病数据集)

shape=(768,9)
该数据集总共九列(特征)：

1. Number of times pregnant 怀孕次数
2. Plasma glucose concentration a 2 hours in an oral glucose tolerance test 葡萄糖
3. Diastolic blood pressure (mm Hg) 血压
4. Triceps skin fold thickness (mm) 皮层厚度
5. 2-Hour serum insulin (mu U/ml) 胰岛素 2小时血清胰岛素
6. Body mass index (weight in kg/(height in m)^2) 体重指数 （体重/身高）
7. Diabetes pedigree function 糖尿病谱系功能
8. Age (years) 年龄 （岁）
9. Class variable (0 or 1) 类标变量 （0或1）是否患病

训练模型

import tensorflow as tf
import numpy
1
2

dataset=numpy.loadtxt('pima-indians-diabetes.csv',delimiter=',')
X=dataset[:,0:8]
Y=dataset[:,8]
1
2
3

#1 define the network
model=tf.keras.Sequential()
model.add(tf.keras.layers.Dense(12,input_dim=8,activation='relu'))
model.add(tf.keras.layers.Dense(1,activation='sigmoid'))

# 2 compile the network 
model.compile(loss='binary_crossentropy',
                             optimizer='adam',
                             metrics=['accuracy']
                            )
#3 fitthe network 
history=model.fit(X,Y,epochs=100,batch_size=10,verbose=1)
1
2
3
4
5
6
7
8
9
10
11
12

最后三个epoch,epoch增多，精度呈现上升趋势上升。

    Epoch 98/100
    768/768 [==============================] - 0s 130us/sample - loss: 0.5204 - accuracy: 0.7591
    Epoch 99/100
    768/768 [==============================] - 0s 126us/sample - loss: 0.5480 - accuracy: 0.7344
    Epoch 100/100
    768/768 [==============================] - 0s 122us/sample - loss: 0.5669 - accuracy: 0.7083
1
2
3
4
5
6
7
8

评估和预测模型

# 4 evaluate the network
loss,accuracy=model.evaluate(X,Y,verbose=1)
print('\nLoss: %.2f, Accuracy: %.2f'%(loss,accuracy*100))

# make predictions
probabilities=model.predict(X)
predictions=[float(numpy.round(x)) for x in probabilities]
acc=numpy.mean(predictions==Y)
print('Prediction Accuracy: %.2f%%'%(acc*100))
1
2
3
4
5
6
7
8
9

768/1 [=====] - 0s 109us/sample - loss: 0.5756 - accuracy: 0.7552

Loss: 0.51, Accuracy: 75.52
Prediction Accuracy: 75.52%
1
2
3
4

保存模型和权重

• 4.6 保存和加载模型 TensorFlow有两种保存模型的方法，一种是只保存模型的权重和偏置，另一种是保存整个模型。基本用法如

model.save_weights('./save_weights/my_save_weigthts')
model.load_weights('./save_weights/my_save_weigthts')
1
2

<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x7f265c603850>
1

model.save('my_model.h5')
restored_model=tf.keras.models.load_model('./my_model.h5')
1
2