用tensorflow预测房价

2.房价影响因素文件(housing.data.csv，housing.data.test.csv)
已上传至资源，请自行下载

3.训练代码实现
①读取数据(data_manager.py)

import pandas as pd

def read_data(path):
    data = pd.read_csv(path)#用panda包读取csv文件
    return data

def get_X_data_Y_data(path):
    All_data=pd.read_csv(path)
    X_data=All_data.iloc[:,:-1]
    Y_data=All_data.iloc[:,-1]  #这里Y_data的类型是pandas里的类型
    Y_data=Y_data.values.reshape(Y_data.shape[0],1)#把Y_data转为数组类型
    return X_data,Y_data

②构建模型(model.py)

import tensorflow as tf

def inference(x,shape):
    weight =tf.Variable(initial_value=tf.truncated_normal(shape=[shape,1]),name='weight')
    bias = tf.Variable(initial_value=tf.truncated_normal(shape=[1]),name='bias')
    y_pred=tf.matmul(x,weight)+bias   #获得预测值
    return y_pred


def losses(y_pred,y):
    loss=tf.reduce_mean(tf.square(y_pred-y),name='loss')
    return loss#获得损失值

def trainning(loss,learning_rate):
    optimizer=tf.train.AdamOptimizer(learning_rate=learning_rate)#使用adam优化器优化
    train_op=optimizer.minimize(loss)#定义优化损失节点
    return train_op

③开始训练(train.py)

import data_manager
import tensorflow as tf
import model
import os


learning_rate=0.01    #学习率
training_epochs=10000 #总训练次数
display_step=100      #每隔100次打印数据
save_step=1000        #每隔1000次保存模型
logs_dir='logs'       #保存模型的目录


def run_train():
    x_data,y_data=data_manager.get_X_data_Y_data('housing.data.csv')
    x_pl=tf.placeholder(tf.float32,shape=x_data.shape,name='X')
    y_pl=tf.placeholder(tf.float32,shape=y_data.shape,name='Y')#定义x_data和y_data的占位符
    y_pred=model.inference(x_pl,x_data.shape[1])
    loss=model.losses(y_pred,y_pl)
    train_op=model.trainning(loss,learning_rate)
    saver=tf.train.Saver()
    with tf.Session() as sess:#创建会话执行图运算
        sess.run(tf.global_variables_initializer())
        for epoch in range(training_epochs):
            _,train_loss=sess.run([train_op,loss],feed_dict={
    
    x_pl:x_data,y_pl:y_data})#注入数据，在没有注入数据之前，是没有训练的
            if epoch%display_step==0 or (epoch+1)==training_epochs:
                print("step:{},train_losses:{}".format(epoch,train_loss))#每隔100次打印训练的损失值
            if epoch%save_step==0 or (epoch+1)==training_epochs:
                checkpoint_path=os.path.join(logs_dir,'model.ckpt')
                saver.save(sess,checkpoint_path,global_step=epoch)#每隔1000次保存模型
                print('step:{},save model'.format(epoch))


def get_wb():
    with tf.Session() as sess:
        saver=tf.train.import_meta_graph('logs/model.ckpt-9999.meta')
        saver.restore(sess,tf.train.latest_checkpoint('logs'))
        w,b=sess.run(['weight:0','bias:0'])#导入模型并打印w,b
        return w,b

run_train()
print(get_wb())

④运行结果

4.计算测试集中的损失值

①代码实现(valid.py)

#计算测试集中的损失值


import tensorflow as tf
import data_manager
import model

logs_dir='logs'

def run_vaild():
    x_data,y_data=data_manager.get_X_data_Y_data("housing.data.test.csv")

    x_pl = tf.placeholder(tf.float32,shape=x_data.shape,name='X')
    y_pl = tf.placeholder(tf.float32,shape=y_data.shape,name='Y')

    y_pred = model.inference(x_pl,x_data.shape[1])
    valid_loss = model.losses(y_pred,y_pl)


    saver=tf.train.Saver()

    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        ckpt = tf.train.latest_checkpoint(logs_dir)
        if ckpt:
            saver.restore(sess,ckpt)
            print("restore from the checkpoint {0}".format(ckpt))
            loss = sess.run([valid_loss], feed_dict={
    
    x_pl: x_data, y_pl: y_data})
            print('valid_losses:{}'.format(loss))

run_vaild()

②结果截图