使用LSTM预测正弦曲线

# -*- coding: utf-8 -*-
"""
Created on Sun Jan  5 21:08:46 2020
@author: Administrator
"""
from __future__ import absolute_import, division, print_function, unicode_literals
import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np
plt.rcParams['font.sans-serif'] = ['SimHei']  # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False  # 用来正常显示负号

def univariate_data(dataset, start_index, end_index, history_size, target_size):
  #一段连续数据做data(长度为history_size)，紧邻的一个数据做label
  #连续滚动,我们就得到了一系列数据和相应的labels
  data = []
  labels = []
  start_index = start_index + history_size
  if end_index is None:
    end_index = len(dataset) - target_size
  for i in range(start_index, end_index):
    indices = range(i-history_size, i)
    # Reshape data from (history_size,) to (history_size, 1)
    data.append(np.reshape(dataset[indices], (history_size, 1)))
    labels.append(dataset[i+target_size])
  return np.array(data), np.array(labels)
  
periods = 1000
points_pp = 16 #sine曲线一个周期取多少个点
X = np.arange(0,periods*2*np.pi-2.0*np.pi/points_pp,2.0*np.pi/points_pp)
Y = np.sin(X)
uni_data = Y
#头12k条数据作为训练集，剩下的4k作为验证集
TRAIN_SPLIT = 12000
#数据标准化（减去均值，再除以标准差）
uni_train_mean = uni_data[:TRAIN_SPLIT].mean()
uni_train_std = uni_data[:TRAIN_SPLIT].std()
uni_data = (uni_data-uni_train_mean)/uni_train_std
univariate_past_history = 48 #用48个历史数据点
univariate_future_target = 16 #预测接下来的16个数据点
x_train_uni, y_train_uni = univariate_data(uni_data, 0, TRAIN_SPLIT,
                                           univariate_past_history,
                                           univariate_future_target)
x_val_uni, y_val_uni = univariate_data(uni_data, TRAIN_SPLIT, None,
                                       univariate_past_history,
                                       univariate_future_target)
BATCH_SIZE = 128 # 128 段 数据
BUFFER_SIZE = 1000
#训练集
tf.random.set_seed(666)
train_univariate = tf.data.Dataset.from_tensor_slices((x_train_uni, y_train_uni))
train_univariate = train_univariate.cache().shuffle(BUFFER_SIZE).batch(BATCH_SIZE).repeat()#打乱训练集
#验证集
val_univariate = tf.data.Dataset.from_tensor_slices((x_val_uni, y_val_uni))
val_univariate = val_univariate.batch(BATCH_SIZE).repeat()

#创建一个简单的LSTM网络模型
simple_lstm_model = tf.keras.models.Sequential([
    tf.keras.layers.LSTM(units=8, input_shape=x_train_uni.shape[-2:],activation="tanh"),#units：输出空间的维度
    tf.keras.layers.Dense(1)
])
simple_lstm_model.compile(optimizer='adam', loss='mae')#模型编译，设定优化器和损失类型

#因为数据集很大，为了节省时间，每个EPOCH仅跑400步，没有跑完所有训练数据
EVALUATION_INTERVAL = 400
EPOCHS = 10
simple_lstm_model.fit(train_univariate, epochs=EPOCHS,
                      steps_per_epoch=EVALUATION_INTERVAL,
                      validation_data=val_univariate, validation_steps=50)

plt.plot(X[-univariate_past_history:],Y[-univariate_past_history:],marker ="o",label ="最后的历史值")
X1 = np.arange(periods*2*np.pi,(periods+1)*2*np.pi - np.pi/8.0 ,np.pi/8.0)
#Y1 = np.sin(X1)
data0 = Y[-univariate_past_history:].copy()
data1 = data0.reshape((1,univariate_past_history,1))
predicts = []
for i in range(univariate_future_target):  
    predict = simple_lstm_model.predict(data1)
    predict = float(predict)
    predicts.append(predict)
    
    #依次将最新的预测值（单个点）作为添加到用于预测的数据的末端，首端弹出最旧的值
    data1[0, 0:-1] = data1[0, 1:]
    data1[0, -1] = predict
plt.plot(X1,predicts,linestyle="--",marker="o",label ="预测值（未来）")
plt.legend(loc="upper right")
plt.title("LSTM sine曲线 预测",fontsize =18)
plt.xlabel('Time')