LSTM-航班人数预测

基于LSTM的航班乘客预测

郑重声明，文章大部分翻译自:
Time Series Prediction with LSTM Recurrent Neural Networks in Python with Keras

数据: 1949 到 1960 一共 12 年，每年 12 个月的数据，一共 144 个数据，单位是 1000

目标: 预测国际航班未来 1 个月的乘客数

#导入相应的库
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 keras.utils import plot_model
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error
from IPython.display import SVG
from keras.utils.vis_utils import model_to_dot

#将数据存储为两个矩阵，一个矩阵的ind位置存储t时刻的值，另一个矩阵存储t+1时刻的值
def create_dataset(dataset, look_back=1):
    dataX, dataY = [], []
    for i in range(len(dataset)-look_back-1):
        a = dataset[i:(i+look_back), 0]
        dataX.append(a)
        dataY.append(dataset[i + look_back, 0])
    return numpy.array(dataX), numpy.array(dataY)

# fix random seed for reproducibility
numpy.random.seed(7)

#读取数据
dataframe = read_csv('international-airline-passengers.csv', usecols=[1], engine='python', skipfooter=3)
dataset = dataframe.values
dataset = dataset.astype('float32')

#查看数据集
print(dataset[0:2],len(dataset))
plt.plot(dataset)
plt.show()

[[112.]
 [118.]] 144

png

#LSTM对输入数据的规模很敏感，特别是在使用sigmoid（默认）或tanh激活函数时。
#将数据重新调整到0到1的范围（也称为标准化）可能是一种很好的做法。

scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(dataset)

# 划分训练集与测试集
train_size = int(len(dataset) * 0.67)
test_size = len(dataset) - train_size
train, test = dataset[0:train_size,:], dataset[train_size:len(dataset),:]

# 生成[t,t+look_back]时间间隔和t+look_back时刻的两个矩阵
look_back = 1
trainX, trainY = create_dataset(train, look_back)
testX, testY = create_dataset(test, look_back)

print(trainX[:2], trainY[:2])

# 数据被Reshape成 [samples, time steps, features]，这是放入LSTM的shape
trainX = numpy.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
testX = numpy.reshape(testX, (testX.shape[0], 1, testX.shape[1]))

print(trainX[:2])

[[0.01544401]
 [0.02702703]] [0.02702703 0.05405405]
[[[0.01544401]]

 [[0.02702703]]]

# 创建并训练LSTM网络
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=50, batch_size=1, verbose=0)

model.summary()

SVG(model_to_dot(model).create(prog='dot', format='svg'))

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
lstm_7 (LSTM)                (None, 4)                 96        
_________________________________________________________________
dense_7 (Dense)              (None, 1)                 5         
=================================================================
Total params: 101
Trainable params: 101
Non-trainable params: 0
_________________________________________________________________

Model

# 使用模型进行预测
trainPredict = model.predict(trainX)
testPredict = model.predict(testX)

# 预测的值是[0,1]，将该值转换回原始值
trainPredict = scaler.inverse_transform(trainPredict)
trainY = scaler.inverse_transform([trainY])
testPredict = scaler.inverse_transform(testPredict)
testY = scaler.inverse_transform([testY])


# 计算预测的均方根误差
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))

# 画图：对训练数据的预测
trainPredictPlot = numpy.empty_like(dataset)
trainPredictPlot[:, :] = numpy.nan
trainPredictPlot[look_back:len(trainPredict)+look_back, :] = trainPredict

# 画图：对测试数据的预测
testPredictPlot = numpy.empty_like(dataset)
testPredictPlot[:, :] = numpy.nan
testPredictPlot[len(trainPredict)+(look_back*2)+1:len(dataset)-1, :] = testPredict

# 显示图片
plt.plot(scaler.inverse_transform(dataset))
plt.plot(trainPredictPlot)
plt.plot(testPredictPlot)
plt.show()

Train Score: 23.40 RMSE
Test Score: 49.64 RMSE

png