【深度学习 项目实战】一个简单的LSTM实现时间序列预测

我们将通过一个简单实例来讲解LSTM建模过程以及模型的使用方法。实例描述如下，我们有一个整数序列[10, 20, 30, 40, 50, 60, 70]，整数序列中共有7个数字，我们的目的或者说我们要解决的问题是，利用已有的这个整数序列来建立模型，用模型预测整数序列的下一个数字应该是多少？也就是整数序列的第8个数字应该是多少？一眼就能看出来，下一个数字应该是80，但是我们需要的是让模型预测出来这个数字。

下面我们梳理一下思路，我们希望向模型输入3个连续的整数，模型能够预测下一个整数，例如，当输入[20, 30, 40]时，模型能够输出[50]。看到这里，想必你已经明白我的意思了，没错，当输入[50, 60, 70]时，我们希望模型能够准确的预测出下一个数字，也就是整数序列的下一个数字。为了训练模型，我们需要把整数序列做一下调整，调整如下：

X表示模型的输入，Y表示模型的输出，每个X,Y对表示一个样本，X表示样本的输入，Y表示样本标签，在本例中共有4个样本。

用Keras建立模型时，要求模型的输入格式为[samples, timesteps, features]（注：samples为样本数，timesteps为时间步数，features为特征数），
我们很期待，当输入[ 50, 60, 70]时，模型能准确预测出80

# LSTM单变量时间序列预测模型
from numpy import array
import numpy as np
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense

# 定义训练样本
X = array([[10, 20, 30], [20, 30, 40], [30, 40, 50], [40, 50, 60]])
y = array([40, 50, 60, 70])

# 将训练样本的数据格式从[samples, timesteps]转换为[samples, timesteps, features]
# X = X.reshape((X.shape[0], X.shape[1], 1))
X = X.reshape((4,3,1))


# 定义模型
model = Sequential()
# 只定义一个LSTM层
model.add(LSTM(50, activation='relu', input_shape=(3, 1)))
# 输出层为全连接层
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse')
# 训练模型
model.fit(X, y, epochs=1500, verbose=0)
# 用模型进行预测，输入值为[50, 60, 70]
x_input = array([50, 60, 70])
x_input = x_input.reshape((1, 3, 1))
yhat = model.predict(x_input, verbose=0)
print(yhat)

Using TensorFlow backend.
2019-12-19 12:18:15.896560: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
2019-12-19 12:18:17.545710: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library nvcuda.dll
2019-12-19 12:18:17.569281: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties: 
name: GeForce GTX 1650 major: 7 minor: 5 memoryClockRate(GHz): 1.56
pciBusID: 0000:01:00.0
2019-12-19 12:18:17.569516: I tensorflow/stream_executor/platform/default/dlopen_checker_stub.cc:25] GPU libraries are statically linked, skip dlopen check.
2019-12-19 12:18:17.570018: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0
2019-12-19 12:18:17.570323: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2019-12-19 12:18:17.572096: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties: 
name: GeForce GTX 1650 major: 7 minor: 5 memoryClockRate(GHz): 1.56
pciBusID: 0000:01:00.0
2019-12-19 12:18:17.572320: I tensorflow/stream_executor/platform/default/dlopen_checker_stub.cc:25] GPU libraries are statically linked, skip dlopen check.
2019-12-19 12:18:17.572812: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0
2019-12-19 12:18:18.096889: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1159] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-12-19 12:18:18.097077: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1165]      0 
2019-12-19 12:18:18.097188: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1178] 0:   N 
2019-12-19 12:18:18.098009: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1304] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 2919 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1650, pci bus id: 0000:01:00.0, compute capability: 7.5)
2019-12-19 12:18:18.986839: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_100.dll
[[80.615295]]

Process finished with exit code 0