数据分析-深度学习-前馈神经网络-分类-Tensorflow

目录

前馈神经

Tensorflow实现-分类

---

前馈神经网络

在前馈神经网络中，每层包含多个神经元，每一层的神经元接受前一层传导的神经元信号，并产生信号输出到下一层。从输入层，经过隐含层（可多层），最后到输出层，整个网络中无反馈，信号从输入层到输出层单向传播。前馈神经网络也被成为多层感知器。

Tensorflow的实现-分类

汽车满意度数据集

import tensorflow as tf
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

读入数据集 

data=pd.read_csv("d:/datasets/car.data",names=["buying","maint","doors","person","lug_boot","safety","car_acc"])

 数据编码

data_=pd.get_dummies(data.drop("car_acc",axis=1))
data_["car_acc"]=data["car_acc"].map({"unacc":1,"acc":0,"good":0,"vgood":0})

 特征标签分离、测试集训练集分离

x=data_.drop("car_acc",axis=1)
y=data_["car_acc"]
from sklearn.model_selection import train_test_split
train_x,test_x,train_y,test_y=train_test_split(x,y,train_size=0.8)

 建立神经网络模型

model = tf.keras.Sequential()  #实例化
model.add(tf.keras.layers.Dense(32,input_shape=(21,),activation="relu")) #增加一层32节点
model.add(tf.keras.layers.Dense(32,activation="relu"))#增加一层32节点 relu激活
model.add(tf.keras.layers.Dense(32,activation="relu"))#增加一层32节点
#可以根据需要增加隐含层
model.add(tf.keras.layers.Dense(1,activation="sigmoid"))  #二分类问题，输出层一个节点，sigmoid激活

model.summary()  #查看模型

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense (Dense)               (None, 32)                704       
                                                                 
 dense_1 (Dense)             (None, 32)                1056      
                                                                 
 dense_2 (Dense)             (None, 32)                1056      
                                                                 
 dense_3 (Dense)             (None, 1)                 33        
                                                                 
=================================================================
Total params: 2,849
Trainable params: 2,849
Non-trainable params: 0

 编译模型

model.compile(optimizer='adam',
              metrics=['acc'],
              loss='binary_crossentropy',  #二分类问题
)

训练模型 

his=model.fit(train_x, 
              train_y,
              batch_size=32,  #每次取32个样本，一个epochs取，样本数/batch_size次
              epochs=50,     #epochs数量
              validation_data=(test_x,test_y)) #验证集

 训练过程可视化

plt.figure(figsize=(14,12))
plt.plot(his.epoch,his.history.get("loss"),label="loss")
plt.plot(his.epoch,his.history.get("val_loss"),label="val_loss")
plt.legend()

  结果评价

from sklearn.metrics import classification_report
y_pre=model.predict(test_x)
y_pre_=[round(x) for i in y_pre for x in i ]
print(classification_report(test_y.values,y_pre_))

              precision    recall  f1-score   support

           0       0.95      1.00      0.98       100
           1       1.00      0.98      0.99       246

    accuracy                           0.99       346
   macro avg       0.98      0.99      0.98       346
weighted avg       0.99      0.99      0.99       346

​