【LSTM】预测

介绍一个棒呆的外国博客网站：https://skymind.ai/wiki/lstm#long
数据集在这
代码在这
介绍一下数据集：

F_day1：前一天value值
F_week: 前一周的当天的value值（今天周一，前一周周一的值）
dayofweek: 周几
isWorkday: 是否工作日
isHoliday: 是否节假日
Tem_max: 温度最高值
Tem_min: 温度最低值
RH_max: 湿度最高值
RH_min: 湿度最低值
（这里用最值是因为给出的是1个小时1个值，按照天计算的话，选了最高和最小值，其他值比较小）
Tag: 大于40000 标1，小于8000为-1，其余为0
kmeans: 温度湿度四个值的聚类分类标签
Value: 用电量
最后的结果图

在这贴个代码

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
import os
#定义常量
rnn_unit= 45 #hidden layer units
input_size=8
output_size=1
lr=0.0001
#导入数据
f=open('./DaySet7.0.csv')
df=pd.read_csv(f)
data=df.iloc[:,1:].values

#获取训练集
def get_train_data(batch_size=60,time_step=30,train_begin=0,train_end=791):
    batch_index=[]
    data_train=data[train_begin:train_end]
    normalized_train_data=(data_train-np.mean(data_train,axis=0))/np.std(data_train,axis=0)
    # print(normalized_train_data)
    train_x,train_y=[],[]
    for i in range(len(normalized_train_data)-time_step):
       if i % batch_size==0:
           batch_index.append(i)
       x=normalized_train_data[i:i+time_step,0:8]
       # print(x)
       y=normalized_train_data[i:i+time_step,8,np.newaxis]
       # print(y)
       train_x.append(x.tolist())
       train_y.append(y.tolist())
    batch_index.append((len(normalized_train_data)-time_step))
    return batch_index,train_x,train_y

#获取测试集
def get_test_data(time_step=30,test_begin=791):
    data_test=data[test_begin:]
    mean=np.mean(data_test,axis=0)
    std=np.std(data_test,axis=0)
    normalized_test_data=(data_test-mean)/std 
    size=(len(normalized_test_data)+time_step-1)//time_step  #有size个sample 
    test_x,test_y=[],[]  
    for i in range(size-1):
       x=normalized_test_data[i*time_step:(i+1)*time_step,:8]
       y=normalized_test_data[i*time_step:(i+1)*time_step,8]
       test_x.append(x.tolist())
       test_y.extend(y)
    return mean,std,test_x,test_y

tf.reset_default_graph()

#——————————————————定义神经网络变量——————————————————
#输入层、输出层权重、偏置

weights={
         'in':tf.Variable(tf.random_normal([input_size,rnn_unit])),
         'out':tf.Variable(tf.random_normal([rnn_unit,1]))
        }
biases={
        'in':tf.Variable(tf.constant(0.1,shape=[rnn_unit,])),
        'out':tf.Variable(tf.constant(0.1,shape=[1,]))
       }

#——————————————————定义神经网络变量——————————————————
def lstm(X):     
    batch_size=tf.shape(X)[0]
    time_step=tf.shape(X)[1]
    w_in=weights['in']
    b_in=biases['in']  
    input=tf.reshape(X,[-1,input_size]) 
    input_rnn=tf.matmul(input,w_in)+b_in
    input_rnn=tf.reshape(input_rnn,[-1,time_step,rnn_unit])
    cell=tf.nn.rnn_cell.BasicLSTMCell(rnn_unit)
    cell = tf.nn.rnn_cell.DropoutWrapper(cell, output_keep_prob=0.8)
    init_state=cell.zero_state(batch_size,dtype=tf.float32)
    output_rnn,final_states=tf.nn.dynamic_rnn(cell, input_rnn,initial_state=init_state, dtype=tf.float32)
    output=tf.reshape(output_rnn,[-1,rnn_unit])
    w_out=weights['out']
    b_out=biases['out']
    pred=tf.matmul(output,w_out)+b_out
    return pred,final_states

#——————————————————训练模型——————————————————
def train_lstm(batch_size=60,time_step=15,train_begin=0,train_end=791):#batch_size=60,time_step=15
    X=tf.placeholder(tf.float32, shape=[None,time_step,input_size])
    Y=tf.placeholder(tf.float32, shape=[None,time_step,output_size])
    batch_index,train_x,train_y=get_train_data(batch_size,time_step,train_begin,train_end)
    pred,_=lstm(X)    
    # loss = tf.sqrt(tf.losses.mean_squared_error(tf.reshape(pred,[-1]),tf.reshape(Y, [-1]))) # rmse
    loss = tf.losses.mean_squared_error(tf.reshape(pred,[-1]),tf.reshape(Y, [-1])) # mse
    train_op = tf.train.AdamOptimizer(lr).minimize(loss)
    saver = tf.train.Saver(tf.global_variables(),max_to_keep=15)#保存最近的15个模型
    #module_file = tf.train.latest_checkpoint() 
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        #saver.restore(sess, module_file)
        #重复训练10000次
        for i in range(100):
            for step in range(len(batch_index)-1):
                _,loss_=sess.run([train_op,loss],feed_dict={X:train_x[batch_index[step]:batch_index[step+1]],Y:train_y[batch_index[step]:batch_index[step+1]]})
            print(i,loss_)
            if i != 0 and i % 20==0:
                print("保存模型：",saver.save(sess,'model_file2' + os.sep+'/stock2.model',global_step=i))


#————————————————预测模型————————————————————
def prediction(time_step=30):
# def prediction(time_step=36):
    X=tf.placeholder(tf.float32, shape=[None,time_step,input_size])
    #Y=tf.placeholder(tf.float32, shape=[None,time_step,output_size])
    mean,std,test_x,test_y=get_test_data(time_step)
    pred,_=lstm(X)     
    saver=tf.train.Saver(tf.global_variables())
    with tf.Session() as sess:
        #参数恢复
        module_file = tf.train.latest_checkpoint('model_file2' + os.sep)
        saver.restore(sess, module_file) 
        test_predict=[]
        for step in range(len(test_x)-1):
            prob=sess.run(pred,feed_dict={X:[test_x[step]]})   
            predict=prob.reshape((-1))
            test_predict.extend(predict)
        test_y=np.array(test_y)*std[8]+mean[8]
        test_predict=np.array(test_predict)*std[8]+mean[8]
        acc=np.average(np.abs(test_predict-test_y[:len(test_predict)])/test_y[:len(test_predict)])  #偏差
        #以折线图表示结果
        plt.figure()
        plt.plot(list(range(len(test_predict))), test_predict, color='b')
        plt.plot(list(range(len(test_y))), test_y,  color='r')
        plt.show()

        MAPE = []
        print(len(test_y),len(test_predict))
        for i in range(len(test_predict)-1):
            print(test_predict[i])
            MAPE.append(np.abs(test_y[i] - test_predict[i]) / test_y[i] * 100)
        MAPE_meanday = np.mean(MAPE)
        print("MAPE:",MAPE_meanday,"%")
        print("sum_test",np.sum(test_y))
        print("sum_pred", np.sum(test_predict))
        print(np.abs(np.sum(test_y) - np.sum(test_predict)) / np.sum(test_y) * 100,"%")

        

with tf.variable_scope('train'):
    train_lstm()

with tf.variable_scope('train',reuse=True):
    prediction()