LSTM进行时间序列预测

示例数据下载

点击此处
目标：预测国际航班未来 1 个月的乘客数
数据如图所示

image.png

import numpy
import matplotlib.pyplot as plt
from pandas import read_csv
import math
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error

导入数据：

# load the dataset
dataframe = read_csv('data/airline-passengers.csv', usecols=[1], engine='python', skipfooter=3)
dataset = dataframe.values
# 将整型变为float
dataset = dataset.astype('float32')
#print(dataframe)
plt.plot(dataset)
#plt.show()
plt.savefig('image/img_show.jpg')

从这 12 年的数据可以看到上升的趋势，每一年内的 12 个月里又有周期性季节性的规律

LSTM进行预测需要的是时序数据根据前timestep步预测后面的数据
假定给一个数据集
{
A,B,C->D
B,C,D->E
C,D,E->F
D,E,F->G
E,F,G->H
}
这时timestep为3，即根据前三个的数据预测后一个数据的值
所以我们需要对数据进行转化
举一个简单的情况假设一个list为[1,2,3,4,5],timestep = 2
我们转化之后要达到的效果是

即依据前两个值预测下一个值

对数据进行归一化
LSTM可以不进行归一化的操作，但是这样会让训练模型的loss下降很慢。本教程如果不进行归一化，100次迭代后loss还是很高

#上面代码的片段讲解
scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(dataset)

对数据进行处理
归一化在下一步会讲解

scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(dataset)
#print(dataset)
train_size = int(len(dataset) * 0.65)
trainlist = dataset[:train_size]
testlist = dataset[train_size:]

print(len(trainlist), len(testlist))

当激活函数为 sigmoid 或者 tanh 时，要把数据正则话，此时 LSTM 比较敏感
设定 67% 是训练数据，余下的是测试数据

# convert an array of values into a dataset matrix
def create_dataset(dataset, look_back=1):
    data_X, data_Y = [], []
    for i in range(len(dataset)-look_back-1):
        a = dataset[i:(i+look_back), 0]
        data_X.append(a) # [0.01544401]
        b = dataset[i + look_back, 0]
        data_Y.append(b) # 0.027027026

    return numpy.array(data_X), numpy.array(data_Y)

# use this function to prepare the train and test datasets for modeling
look_back = 1
trainX, trainY = create_dataset(trainlist, look_back)
testX, testY = create_dataset(testlist, look_back)

X=t and Y=t+1 时的数据，并且此时的维度为 [samples, features]

look_back = 1
trainX, trainY = create_dataset(trainlist, look_back)
testX, testY = create_dataset(testlist, look_back)

投入到 LSTM 的 X 需要有这样的结构： [samples, time steps, features]，所以做一下变换

# reshape input to be [samples, time steps, features]
trainX = numpy.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
testX = numpy.reshape(testX, (testX.shape[0], 1, testX.shape[1]))

[[[0.6891892 ]]
 [[0.71042466]]
.....
 [[0.8320464 ]]
 [[1.        ]]]

建立 LSTM 模型：
输入层有 1 个input，隐藏层有 4 个神经元，输出层就是预测一个值，激活函数用 sigmoid，迭代 100 次，batch size 为 1

# create and fit the LSTM network
model = Sequential()
model.add(LSTM(4, input_shape=(1, look_back)))
model.add(Dense(1))
model.compile(loss='mean_squared_error', optimizer='adam')
model.fit(trainX, trainY, epochs=100, batch_size=1, verbose=2)
model.save(os.path.join("data","Test" + ".h5"))

Epoch 100/100

0s - loss: 0.0018
进行预测

model = load_model(os.path.join("data","Test" + ".h5"))
trainPredict = model.predict(trainX)
testPredict = model.predict(testX)

计算误差之前要先把预测数据转换成同一单位
反归一化

trainPredict = scaler.inverse_transform(trainPredict)
trainY = scaler.inverse_transform([trainY])
testPredict = scaler.inverse_transform(testPredict)
testY = scaler.inverse_transform([testY])

计算 mean squared error

trainScore = math.sqrt(mean_squared_error(trainY[0], trainPredict[:,0]))
print('Train Score: %.2f RMSE' % (trainScore))
testScore = math.sqrt(mean_squared_error(testY[0], testPredict[:,0]))
print('Test Score: %.2f RMSE' % (testScore))

Train Score: 21.75 RMSE
Test Score: 46.30 RMSE
画出结果：蓝色为原数据，绿色为训练集的预测值，红色为测试集的预测值

# shift train predictions for plotting
trainPredictPlot = numpy.empty_like(dataset)
trainPredictPlot[:, :] = numpy.nan
trainPredictPlot[look_back:len(trainPredict)+look_back, :] = trainPredict

# shift test predictions for plotting
testPredictPlot = numpy.empty_like(dataset)
testPredictPlot[:, :] = numpy.nan
testPredictPlot[len(trainPredict)+(look_back*2)+1:len(dataset)-1, :] = testPredict

# plot baseline and predictions
plt.plot(scaler.inverse_transform(dataset))
plt.plot(trainPredictPlot)
plt.plot(testPredictPlot)
plt.show()