《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (12): 3931-3940.DOI: 10.11772/j.issn.1001-9081.2021101716
收稿日期:
2021-10-08
修回日期:
2022-01-04
接受日期:
2022-01-24
发布日期:
2022-04-08
出版日期:
2022-12-10
通讯作者:
栗慧琳
作者简介:
李洪涛(1976—),男,河南漯河人,教授,博士,主要研究方向:机器学习、预测理论与方法、预测与决策基金资助:
Huilin LI(), Hongtao LI, Zhi LI
Received:
2021-10-08
Revised:
2022-01-04
Accepted:
2022-01-24
Online:
2022-04-08
Published:
2022-12-10
Contact:
Huilin LI
About author:
LI Hongtao born in 1976, Ph. D., professor. His research interests include machine learning, prediction theory and method,prediction and decision.Supported by:
摘要:
考虑到航空客流需求序列的季节性、非线性和非平稳等特点,提出了一个基于二次分解重构策略的航空客流需求预测模型。首先,通过STL和自适应噪声互补集成经验模态分解(CEEMDAN)方法对航空客流需求序列进行二次分解,并根据数据复杂度和相关度的特征分析结果进行分量重构;然后,采用模型匹配策略分别选取自回归单整移动平均季节(SARIMA)、自回归单整移动平均(ARIMA)、核极限学习机(KELM)和双向长短期记忆(BiLSTM)网络模型对各重构分量进行预测,其中KELM和BiLSTM模型的超参数通过自适应树Parzen估计(ATPE)算法确定;最后,将重构分量预测结果进行线性集成。以北京首都国际机场、深圳宝安国际机场和海口美兰国际机场的航空客流数据作为研究对象进行了1步和多步预测实验,实验结果表明,与一次分解集成模型STL-SAAB相比,所提模型的均方根误差(RMSE)提升了14.98%~60.72%。可见以“分而治之”思想为指导,所提模型结合模型匹配和重构策略挖掘出了数据的内在发展规律,从而为科学预判航空客流需求变化趋势提供了新思路。
中图分类号:
栗慧琳, 李洪涛, 李智. 基于二次分解重构策略的航空客流需求预测[J]. 计算机应用, 2022, 42(12): 3931-3940.
Huilin LI, Hongtao LI, Zhi LI. Air passenger demand forecasting based on dual decomposition and reconstruction strategy[J]. Journal of Computer Applications, 2022, 42(12): 3931-3940.
机场 | 最小值 | 最大值 | 均值 | 标准差 | 偏度 | 峰度 |
---|---|---|---|---|---|---|
北京 | 316.14 | 904.50 | 659.83 | 146.55 | -0.46 | 2.11 |
深圳 | 138.83 | 469.12 | 278.19 | 90.57 | 0.35 | 1.94 |
海口 | 39.09 | 251.86 | 114.21 | 55.33 | 0.71 | 2.28 |
表1 3个机场航空客流需求月度数据的基本统计描述 (万人次)
Tab. 1 Basic description statistics of monthly air passenger demand data of three airports
机场 | 最小值 | 最大值 | 均值 | 标准差 | 偏度 | 峰度 |
---|---|---|---|---|---|---|
北京 | 316.14 | 904.50 | 659.83 | 146.55 | -0.46 | 2.11 |
深圳 | 138.83 | 469.12 | 278.19 | 90.57 | 0.35 | 1.94 |
海口 | 39.09 | 251.86 | 114.21 | 55.33 | 0.71 | 2.28 |
评价指标 | 定义 | 公式 |
---|---|---|
RMSE | 均方根误差 | |
MAPE | 平均绝对百分比误差 | |
R 2 | 决定系数 | |
DA | 方向预测精度 |
表2 评价指标
Tab. 2 Evaluation indicators
评价指标 | 定义 | 公式 |
---|---|---|
RMSE | 均方根误差 | |
MAPE | 平均绝对百分比误差 | |
R 2 | 决定系数 | |
DA | 方向预测精度 |
分量 | CEEMDAN | EMD | ||
---|---|---|---|---|
MIC | SE | MIC | SE | |
IMF1 | 0.016 9 | 1.413 7 | 0.064 4 | 1.107 0 |
IMF2 | 0.022 3 | 0.392 3 | 0.046 3 | 0.703 1 |
IMF3 | 0.038 4 | 0.709 7 | 0.070 7 | 0.561 8 |
IMF4 | 0.104 5 | 0.645 6 | 0.228 1 | 0.615 8 |
IMF5 | 0.168 2 | 0.439 6 | 0.305 4 | 0.277 7 |
IMF6 | 0.140 2 | 0.412 3 | 0.187 2 | 0.362 2 |
IMF7 | 0.458 1 | 0.141 0 | 0.776 7 | 0.084 8 |
IMF8 | 1.000 0 | 0.027 5 | 1.000 0 | 0.012 5 |
表3 二次分解分量的基本特征统计值
Tab. 3 Basic characteristic statistical values of dual decomposition components
分量 | CEEMDAN | EMD | ||
---|---|---|---|---|
MIC | SE | MIC | SE | |
IMF1 | 0.016 9 | 1.413 7 | 0.064 4 | 1.107 0 |
IMF2 | 0.022 3 | 0.392 3 | 0.046 3 | 0.703 1 |
IMF3 | 0.038 4 | 0.709 7 | 0.070 7 | 0.561 8 |
IMF4 | 0.104 5 | 0.645 6 | 0.228 1 | 0.615 8 |
IMF5 | 0.168 2 | 0.439 6 | 0.305 4 | 0.277 7 |
IMF6 | 0.140 2 | 0.412 3 | 0.187 2 | 0.362 2 |
IMF7 | 0.458 1 | 0.141 0 | 0.776 7 | 0.084 8 |
IMF8 | 1.000 0 | 0.027 5 | 1.000 0 | 0.012 5 |
参数 | 值 |
---|---|
SARIMA | (4,0,4)(2,1,2)12 |
ARIMA | (3,2,2) |
KELM核函数 | 线性核函数 |
KELM惩罚参数 C | 14.999 3 |
BiLSTM学习率 | 0.010 2 |
BiLSTM批处理大小 | 38 |
BiLSTM隐藏层数 | 3 |
BiLSTM隐藏层单元数 | 32 |
表4 各模型参数及超参数
Tab. 4 Parameters and hyperparameters of different models
参数 | 值 |
---|---|
SARIMA | (4,0,4)(2,1,2)12 |
ARIMA | (3,2,2) |
KELM核函数 | 线性核函数 |
KELM惩罚参数 C | 14.999 3 |
BiLSTM学习率 | 0.010 2 |
BiLSTM批处理大小 | 38 |
BiLSTM隐藏层数 | 3 |
BiLSTM隐藏层单元数 | 32 |
预测步长 | 模型 | RMSE | MAPE/% | R 2 | DA |
---|---|---|---|---|---|
1步 | ARIMA | 38.215 6 | 3.848 6 | 0.037 5 | 0.645 2 |
SARIMA | 30.788 4 | 2.665 8 | 0.375 3 | 0.806 5 | |
AK | 31.265 6 | 3.216 3 | 0.355 8 | 0.838 7 | |
AB | 27.961 3 | 2.807 5 | 0.484 8 | 0.774 2 | |
STL-AK | 28.342 8 | 2.410 0 | 0.470 6 | 0.774 2 | |
STL-AB | 16.975 8 | 1.724 2 | 0.810 1 | 0.871 0 | |
STL-SAAB | 14.596 6 | 1.153 6 | 0.859 6 | 0.903 2 | |
STL-EMD-SAAKAB | 7.346 8 | 0.673 7 | 0.964 4 | 0.967 7 | |
本文模型 | 5.733 6 | 0.512 6 | 0.978 3 | 0.935 5 | |
2步 | ARIMA | 38.755 2 | 4.164 8 | 0.016 8 | 0.709 7 |
SARIMA | 35.727 5 | 3.384 2 | 0.164 4 | 0.806 5 | |
AK | 34.306 5 | 3.529 2 | 0.229 5 | 0.838 7 | |
AB | 27.845 7 | 2.802 4 | 0.492 4 | 0.838 7 | |
STL-AK | 29.798 7 | 2.556 8 | 0.418 7 | 0.838 7 | |
STL-AB | 19.159 4 | 1.863 4 | 0.759 7 | 0.838 7 | |
STL-SAAB | 17.535 2 | 1.471 4 | 0.798 7 | 0.871 0 | |
STL-EMD-SAAKAB | 10.059 8 | 0.853 6 | 0.933 8 | 0.935 5 | |
本文模型 | 9.424 5 | 0.786 9 | 0.941 9 | 0.935 5 | |
3步 | ARIMA | 38.679 6 | 4.091 1 | 0.013 8 | 0.645 2 |
SARIMA | 32.015 8 | 3.252 8 | 0.122 4 | 0.741 9 | |
AK | 32.805 7 | 3.292 8 | 0.078 5 | 0.838 7 | |
AB | 28.038 4 | 2.903 9 | 0.326 9 | 0.806 5 | |
STL-AK | 24.751 7 | 2.184 0 | 0.475 5 | 0.774 2 | |
STL-AB | 16.948 2 | 1.558 5 | 0.754 1 | 0.935 5 | |
STL-SAAB | 17.436 5 | 1.615 5 | 0.739 7 | 0.903 2 | |
STL-EMD-SAAKAB | 9.627 1 | 0.946 9 | 0.920 6 | 0.967 7 | |
本文模型 | 12.524 4 | 1.229 0 | 0.865 7 | 0.903 2 |
表5 各模型的1步、2步、3步预测结果(北京首都国际机场)
Tab.5 One step, 2 step and 3 step ahead prediction results of different models (Beijing Capital International Airport)
预测步长 | 模型 | RMSE | MAPE/% | R 2 | DA |
---|---|---|---|---|---|
1步 | ARIMA | 38.215 6 | 3.848 6 | 0.037 5 | 0.645 2 |
SARIMA | 30.788 4 | 2.665 8 | 0.375 3 | 0.806 5 | |
AK | 31.265 6 | 3.216 3 | 0.355 8 | 0.838 7 | |
AB | 27.961 3 | 2.807 5 | 0.484 8 | 0.774 2 | |
STL-AK | 28.342 8 | 2.410 0 | 0.470 6 | 0.774 2 | |
STL-AB | 16.975 8 | 1.724 2 | 0.810 1 | 0.871 0 | |
STL-SAAB | 14.596 6 | 1.153 6 | 0.859 6 | 0.903 2 | |
STL-EMD-SAAKAB | 7.346 8 | 0.673 7 | 0.964 4 | 0.967 7 | |
本文模型 | 5.733 6 | 0.512 6 | 0.978 3 | 0.935 5 | |
2步 | ARIMA | 38.755 2 | 4.164 8 | 0.016 8 | 0.709 7 |
SARIMA | 35.727 5 | 3.384 2 | 0.164 4 | 0.806 5 | |
AK | 34.306 5 | 3.529 2 | 0.229 5 | 0.838 7 | |
AB | 27.845 7 | 2.802 4 | 0.492 4 | 0.838 7 | |
STL-AK | 29.798 7 | 2.556 8 | 0.418 7 | 0.838 7 | |
STL-AB | 19.159 4 | 1.863 4 | 0.759 7 | 0.838 7 | |
STL-SAAB | 17.535 2 | 1.471 4 | 0.798 7 | 0.871 0 | |
STL-EMD-SAAKAB | 10.059 8 | 0.853 6 | 0.933 8 | 0.935 5 | |
本文模型 | 9.424 5 | 0.786 9 | 0.941 9 | 0.935 5 | |
3步 | ARIMA | 38.679 6 | 4.091 1 | 0.013 8 | 0.645 2 |
SARIMA | 32.015 8 | 3.252 8 | 0.122 4 | 0.741 9 | |
AK | 32.805 7 | 3.292 8 | 0.078 5 | 0.838 7 | |
AB | 28.038 4 | 2.903 9 | 0.326 9 | 0.806 5 | |
STL-AK | 24.751 7 | 2.184 0 | 0.475 5 | 0.774 2 | |
STL-AB | 16.948 2 | 1.558 5 | 0.754 1 | 0.935 5 | |
STL-SAAB | 17.436 5 | 1.615 5 | 0.739 7 | 0.903 2 | |
STL-EMD-SAAKAB | 9.627 1 | 0.946 9 | 0.920 6 | 0.967 7 | |
本文模型 | 12.524 4 | 1.229 0 | 0.865 7 | 0.903 2 |
模型 | 1步 | 2步 | 3步 | |||
---|---|---|---|---|---|---|
DM | p值 | DM | p值 | DM | p值 | |
STL-EMD- SAAKAB | -1.503 0 | 0.143 0 | -0.4552 | 0.652 2 | -1.392 0 | 0.173 8 |
STL-SAAB | -2.038 3 | 0.050 1 | -1.7499 | 0.090 0 | -1.749 0 | 0.090 2 |
STL-AB | -3.572 7 | 0.001 2 | -2.8116 | 0.008 5 | -1.872 2 | 0.084 5 |
STL-AK | -2.688 5 | 0.011 4 | -2.5249 | 0.016 9 | -2.122 4 | 0.041 9 |
AB | -4.498 5 | 0.000 1 | -3.6735 | 0.000 9 | -10.904 9 | 0.000 0 |
AK | -4.532 6 | 0.000 1 | -4.6335 | 0.000 1 | -3.382 3 | 0.002 0 |
SARIMA | -2.915 1 | 0.006 5 | -3.2303 | 0.002 9 | -4.062 2 | 0.000 3 |
ARIMA | -5.010 2 | 0.000 0 | -5.2175 | 0.000 0 | -4.459 2 | 0.000 1 |
表6 各模型1步、2步和3步预测的DM检验结果(北京首都国际机场)
Tab.6 DM test results of 1 step, 2 step and 3 step ahead prediction of different models (Beijing Capital International Airport)
模型 | 1步 | 2步 | 3步 | |||
---|---|---|---|---|---|---|
DM | p值 | DM | p值 | DM | p值 | |
STL-EMD- SAAKAB | -1.503 0 | 0.143 0 | -0.4552 | 0.652 2 | -1.392 0 | 0.173 8 |
STL-SAAB | -2.038 3 | 0.050 1 | -1.7499 | 0.090 0 | -1.749 0 | 0.090 2 |
STL-AB | -3.572 7 | 0.001 2 | -2.8116 | 0.008 5 | -1.872 2 | 0.084 5 |
STL-AK | -2.688 5 | 0.011 4 | -2.5249 | 0.016 9 | -2.122 4 | 0.041 9 |
AB | -4.498 5 | 0.000 1 | -3.6735 | 0.000 9 | -10.904 9 | 0.000 0 |
AK | -4.532 6 | 0.000 1 | -4.6335 | 0.000 1 | -3.382 3 | 0.002 0 |
SARIMA | -2.915 1 | 0.006 5 | -3.2303 | 0.002 9 | -4.062 2 | 0.000 3 |
ARIMA | -5.010 2 | 0.000 0 | -5.2175 | 0.000 0 | -4.459 2 | 0.000 1 |
预测步长 | 模型 | RMSE | MAPE/% | R 2 | DA |
---|---|---|---|---|---|
1步 | ARIMA | 17.065 2 | 3.347 1 | 0.625 5 | 0.741 9 |
SARIMA | 10.823 3 | 2.107 4 | 0.849 4 | 0.838 7 | |
AK | 12.183 5 | 2.495 7 | 0.809 1 | 0.871 0 | |
AB | 17.100 1 | 3.447 3 | 0.624 0 | 0.741 9 | |
STL-AK | 8.572 0 | 1.568 9 | 0.905 5 | 0.838 7 | |
STL-AB | 3.884 1 | 0.829 9 | 0.980 6 | 0.903 2 | |
STL-SAAB | 2.823 6 | 0.465 0 | 0.989 7 | 0.935 5 | |
STL-EMD-SAAKAB | 2.561 2 | 0.407 9 | 0.991 6 | 0.935 5 | |
本文模型 | 2.322 5 | 0.438 4 | 0.993 1 | 0.935 5 | |
2步 | ARIMA | 16.415 4 | 3.154 0 | 0.649 6 | 0.774 2 |
SARIMA | 10.939 7 | 2.132 4 | 0.844 4 | 0.774 2 | |
AK | 11.732 3 | 2.306 5 | 0.821 0 | 0.838 7 | |
AB | 17.805 8 | 3.463 4 | 0.587 7 | 0.709 7 | |
STL-AK | 8.581 4 | 1.593 2 | 0.904 2 | 0.838 7 | |
STL-AB | 5.281 8 | 0.945 9 | 0.963 7 | 0.935 5 | |
STL-SAAB | 4.222 1 | 0.732 6 | 0.976 8 | 0.903 2 | |
STL-EMD-SAAKAB | 2.901 6 | 0.522 9 | 0.989 1 | 0.935 5 | |
本文模型 | 2.858 9 | 0.506 1 | 0.989 4 | 0.935 5 | |
3步 | ARIMA | 16.762 8 | 3.366 6 | 0.628 6 | 0.741 9 |
SARIMA | 11.303 5 | 2.280 2 | 0.831 1 | 0.903 2 | |
AK | 11.905 9 | 2.346 8 | 0.812 6 | 0.935 5 | |
AB | 21.996 7 | 4.532 0 | 0.360 4 | 0.677 4 | |
STL-AK | 8.529 9 | 1.594 8 | 0.903 8 | 0.871 0 | |
STL-AB | 6.042 1 | 1.105 0 | 0.951 7 | 0.935 5 | |
STL-SAAB | 5.434 5 | 1.033 7 | 0.961 0 | 0.903 2 | |
STL-EMD-SAAKAB | 3.961 3 | 0.719 1 | 0.979 3 | 0.903 2 | |
本文模型 | 3.309 8 | 0.624 5 | 0.985 5 | 0.935 5 |
表7 各模型的1步、2步、3步预测结果(深圳宝安国际机场)
Tab.7 One step, 2 step and 3 step ahead prediction results of different models (Shenzhen Bao’an International Airport)
预测步长 | 模型 | RMSE | MAPE/% | R 2 | DA |
---|---|---|---|---|---|
1步 | ARIMA | 17.065 2 | 3.347 1 | 0.625 5 | 0.741 9 |
SARIMA | 10.823 3 | 2.107 4 | 0.849 4 | 0.838 7 | |
AK | 12.183 5 | 2.495 7 | 0.809 1 | 0.871 0 | |
AB | 17.100 1 | 3.447 3 | 0.624 0 | 0.741 9 | |
STL-AK | 8.572 0 | 1.568 9 | 0.905 5 | 0.838 7 | |
STL-AB | 3.884 1 | 0.829 9 | 0.980 6 | 0.903 2 | |
STL-SAAB | 2.823 6 | 0.465 0 | 0.989 7 | 0.935 5 | |
STL-EMD-SAAKAB | 2.561 2 | 0.407 9 | 0.991 6 | 0.935 5 | |
本文模型 | 2.322 5 | 0.438 4 | 0.993 1 | 0.935 5 | |
2步 | ARIMA | 16.415 4 | 3.154 0 | 0.649 6 | 0.774 2 |
SARIMA | 10.939 7 | 2.132 4 | 0.844 4 | 0.774 2 | |
AK | 11.732 3 | 2.306 5 | 0.821 0 | 0.838 7 | |
AB | 17.805 8 | 3.463 4 | 0.587 7 | 0.709 7 | |
STL-AK | 8.581 4 | 1.593 2 | 0.904 2 | 0.838 7 | |
STL-AB | 5.281 8 | 0.945 9 | 0.963 7 | 0.935 5 | |
STL-SAAB | 4.222 1 | 0.732 6 | 0.976 8 | 0.903 2 | |
STL-EMD-SAAKAB | 2.901 6 | 0.522 9 | 0.989 1 | 0.935 5 | |
本文模型 | 2.858 9 | 0.506 1 | 0.989 4 | 0.935 5 | |
3步 | ARIMA | 16.762 8 | 3.366 6 | 0.628 6 | 0.741 9 |
SARIMA | 11.303 5 | 2.280 2 | 0.831 1 | 0.903 2 | |
AK | 11.905 9 | 2.346 8 | 0.812 6 | 0.935 5 | |
AB | 21.996 7 | 4.532 0 | 0.360 4 | 0.677 4 | |
STL-AK | 8.529 9 | 1.594 8 | 0.903 8 | 0.871 0 | |
STL-AB | 6.042 1 | 1.105 0 | 0.951 7 | 0.935 5 | |
STL-SAAB | 5.434 5 | 1.033 7 | 0.961 0 | 0.903 2 | |
STL-EMD-SAAKAB | 3.961 3 | 0.719 1 | 0.979 3 | 0.903 2 | |
本文模型 | 3.309 8 | 0.624 5 | 0.985 5 | 0.935 5 |
模型 | 1步 | 2步 | 3步 | |||
---|---|---|---|---|---|---|
DM | p值 | DM | p值 | DM | p值 | |
STL-EMD- SAAKAB | -0.428 2 | 0.671 5 | -0.085 5 | 0.932 4 | -0.751 1 | 0.458 3 |
STL-SAAB | -1.258 7 | 0.217 5 | -2.480 7 | 0.018 7 | -3.744 9 | 0.000 7 |
STL-AB | -5.909 6 | 0.000 0 | -3.760 8 | 0.000 7 | -4.154 9 | 0.000 2 |
STL-AK | -4.170 4 | 0.000 2 | -4.434 0 | 0.000 1 | -4.307 2 | 0.000 2 |
AB | -7.414 3 | 0.000 0 | -6.340 1 | 0.000 0 | -7.823 1 | 0.000 0 |
AK | -5.928 4 | 0.000 0 | -7.630 5 | 0.000 0 | -8.437 5 | 0.000 0 |
SARIMA | -7.249 4 | 0.000 0 | -8.562 5 | 0.000 0 | -7.052 5 | 0.000 0 |
ARIMA | -6.635 4 | 0.000 0 | -7.755 3 | 0.000 0 | -9.629 5 | 0.000 0 |
表8 各模型1步、2步和3步预测的DM检验结果(深圳宝安国际机场)
Tab.8 DM test results of 1 step, 2 step and 3 step ahead prediction of different models (Shenzhen Bao’an International Airport)
模型 | 1步 | 2步 | 3步 | |||
---|---|---|---|---|---|---|
DM | p值 | DM | p值 | DM | p值 | |
STL-EMD- SAAKAB | -0.428 2 | 0.671 5 | -0.085 5 | 0.932 4 | -0.751 1 | 0.458 3 |
STL-SAAB | -1.258 7 | 0.217 5 | -2.480 7 | 0.018 7 | -3.744 9 | 0.000 7 |
STL-AB | -5.909 6 | 0.000 0 | -3.760 8 | 0.000 7 | -4.154 9 | 0.000 2 |
STL-AK | -4.170 4 | 0.000 2 | -4.434 0 | 0.000 1 | -4.307 2 | 0.000 2 |
AB | -7.414 3 | 0.000 0 | -6.340 1 | 0.000 0 | -7.823 1 | 0.000 0 |
AK | -5.928 4 | 0.000 0 | -7.630 5 | 0.000 0 | -8.437 5 | 0.000 0 |
SARIMA | -7.249 4 | 0.000 0 | -8.562 5 | 0.000 0 | -7.052 5 | 0.000 0 |
ARIMA | -6.635 4 | 0.000 0 | -7.755 3 | 0.000 0 | -9.629 5 | 0.000 0 |
预测步长 | 模型 | RMSE | MAPE/% | R 2 | DA |
---|---|---|---|---|---|
1步 | ARIMA | 18.516 1 | 7.896 0 | 0.439 9 | 0.548 4 |
SARIMA | 8.506 4 | 3.527 8 | 0.881 8 | 1.000 0 | |
AK | 11.820 1 | 5.170 3 | 0.771 8 | 0.838 7 | |
AB | 15.145 8 | 6.198 8 | 0.625 2 | 0.774 2 | |
STL-AK | 6.615 5 | 2.629 8 | 0.928 5 | 1.000 0 | |
STL-AB | 10.542 0 | 4.615 1 | 0.818 4 | 0.871 0 | |
STL-SAAB | 6.685 2 | 2.445 2 | 0.927 0 | 1.000 0 | |
STL-EMD-SAAKAB | 3.080 4 | 1.339 9 | 0.984 5 | 1.000 0 | |
本文模型 | 2.690 5 | 1.166 1 | 0.988 2 | 1.000 0 | |
2步 | ARIMA | 26.537 2 | 11.435 5 | 0.165 3 | 0.451 6 |
SARIMA | 10.384 5 | 4.738 7 | 0.823 4 | 0.871 0 | |
AK | 12.239 5 | 5.531 4 | 0.754 6 | 0.871 0 | |
AB | 21.311 9 | 8.476 2 | 0.256 0 | 0.741 9 | |
STL-AK | 7.584 6 | 2.969 2 | 0.905 8 | 1.000 0 | |
STL-AB | 8.462 6 | 3.690 2 | 0.882 7 | 0.935 5 | |
STL-SAAB | 7.176 8 | 2.632 6 | 0.915 6 | 1.000 0 | |
STL-EMD-SAAKAB | 6.241 3 | 2.730 5 | 0.936 2 | 0.967 7 | |
本文模型 | 5.431 9 | 2.267 4 | 0.951 7 | 1.000 0 | |
3步 | ARIMA | 29.087 5 | 12.235 8 | 0.123 2 | 0.516 1 |
SARIMA | 12.304 4 | 5.607 1 | 0.751 5 | 0.871 0 | |
AK | 13.043 2 | 5.917 7 | 0.720 8 | 0.838 7 | |
AB | 21.929 5 | 9.406 5 | 0.210 6 | 0.741 9 | |
STL-AK | 8.844 6 | 3.619 6 | 0.871 6 | 0.967 7 | |
STL-AB | 8.492 7 | 3.285 5 | 0.881 6 | 0.967 7 | |
STL-SAAB | 8.134 4 | 3.369 8 | 0.891 4 | 1.000 0 | |
STL-EMD-SAAKAB | 7.406 5 | 2.779 5 | 0.910 0 | 0.967 7 | |
本文模型 | 6.916 2 | 2.633 4 | 0.921 5 | 0.967 7 |
表9 各模型的1步、2步、3步预测结果(海口美兰国际机场)
Tab.9 One step, 2 step and 3 step ahead prediction results of different models (Haikou Meilan International Airport)
预测步长 | 模型 | RMSE | MAPE/% | R 2 | DA |
---|---|---|---|---|---|
1步 | ARIMA | 18.516 1 | 7.896 0 | 0.439 9 | 0.548 4 |
SARIMA | 8.506 4 | 3.527 8 | 0.881 8 | 1.000 0 | |
AK | 11.820 1 | 5.170 3 | 0.771 8 | 0.838 7 | |
AB | 15.145 8 | 6.198 8 | 0.625 2 | 0.774 2 | |
STL-AK | 6.615 5 | 2.629 8 | 0.928 5 | 1.000 0 | |
STL-AB | 10.542 0 | 4.615 1 | 0.818 4 | 0.871 0 | |
STL-SAAB | 6.685 2 | 2.445 2 | 0.927 0 | 1.000 0 | |
STL-EMD-SAAKAB | 3.080 4 | 1.339 9 | 0.984 5 | 1.000 0 | |
本文模型 | 2.690 5 | 1.166 1 | 0.988 2 | 1.000 0 | |
2步 | ARIMA | 26.537 2 | 11.435 5 | 0.165 3 | 0.451 6 |
SARIMA | 10.384 5 | 4.738 7 | 0.823 4 | 0.871 0 | |
AK | 12.239 5 | 5.531 4 | 0.754 6 | 0.871 0 | |
AB | 21.311 9 | 8.476 2 | 0.256 0 | 0.741 9 | |
STL-AK | 7.584 6 | 2.969 2 | 0.905 8 | 1.000 0 | |
STL-AB | 8.462 6 | 3.690 2 | 0.882 7 | 0.935 5 | |
STL-SAAB | 7.176 8 | 2.632 6 | 0.915 6 | 1.000 0 | |
STL-EMD-SAAKAB | 6.241 3 | 2.730 5 | 0.936 2 | 0.967 7 | |
本文模型 | 5.431 9 | 2.267 4 | 0.951 7 | 1.000 0 | |
3步 | ARIMA | 29.087 5 | 12.235 8 | 0.123 2 | 0.516 1 |
SARIMA | 12.304 4 | 5.607 1 | 0.751 5 | 0.871 0 | |
AK | 13.043 2 | 5.917 7 | 0.720 8 | 0.838 7 | |
AB | 21.929 5 | 9.406 5 | 0.210 6 | 0.741 9 | |
STL-AK | 8.844 6 | 3.619 6 | 0.871 6 | 0.967 7 | |
STL-AB | 8.492 7 | 3.285 5 | 0.881 6 | 0.967 7 | |
STL-SAAB | 8.134 4 | 3.369 8 | 0.891 4 | 1.000 0 | |
STL-EMD-SAAKAB | 7.406 5 | 2.779 5 | 0.910 0 | 0.967 7 | |
本文模型 | 6.916 2 | 2.633 4 | 0.921 5 | 0.967 7 |
模型 | 1步 | 2步 | 3步 | |||
---|---|---|---|---|---|---|
DM | p值 | DM | p值 | DM | p值 | |
STL-EMD- SAAKAB | -1.485 1 | 0.147 6 | -1.511 7 | 0.140 7 | -0.8811 | 0.385 1 |
STL-SAAB | -2.200 8 | 0.035 3 | -1.863 2 | 0.073 6 | -1.7473 | 0.097 8 |
STL-AB | -4.858 4 | 0.000 0 | -4.995 9 | 0.000 0 | -1.8379 | 0.075 7 |
STL-AK | -3.284 7 | 0.002 5 | -1.903 9 | 0.062 7 | -2.0238 | 0.043 9 |
AB | -4.246 6 | 0.000 2 | -2.178 8 | 0.037 1 | -2.4305 | 0.021 1 |
AK | -5.383 6 | 0.000 0 | -3.694 1 | 0.000 8 | -3.0034 | 0.005 2 |
SARIMA | -4.451 2 | 0.000 1 | -4.308 3 | 0.000 2 | -4.3382 | 0.000 1 |
ARIMA | -7.763 4 | 0.000 0 | -8.400 7 | 0.000 0 | -4.8758 | 0.000 0 |
表10 各模型1步、2步和3步预测的DM检验结果(海口美兰国际机场)
Tab.10 DM test results of 1 step, 2 step and 3 step ahead prediction of different models (Haikou Meilan International Airport)
模型 | 1步 | 2步 | 3步 | |||
---|---|---|---|---|---|---|
DM | p值 | DM | p值 | DM | p值 | |
STL-EMD- SAAKAB | -1.485 1 | 0.147 6 | -1.511 7 | 0.140 7 | -0.8811 | 0.385 1 |
STL-SAAB | -2.200 8 | 0.035 3 | -1.863 2 | 0.073 6 | -1.7473 | 0.097 8 |
STL-AB | -4.858 4 | 0.000 0 | -4.995 9 | 0.000 0 | -1.8379 | 0.075 7 |
STL-AK | -3.284 7 | 0.002 5 | -1.903 9 | 0.062 7 | -2.0238 | 0.043 9 |
AB | -4.246 6 | 0.000 2 | -2.178 8 | 0.037 1 | -2.4305 | 0.021 1 |
AK | -5.383 6 | 0.000 0 | -3.694 1 | 0.000 8 | -3.0034 | 0.005 2 |
SARIMA | -4.451 2 | 0.000 1 | -4.308 3 | 0.000 2 | -4.3382 | 0.000 1 |
ARIMA | -7.763 4 | 0.000 0 | -8.400 7 | 0.000 0 | -4.8758 | 0.000 0 |
1 | CHAN L L, GULERIA Y, ALAM S. Air passenger forecasting using Neural Granger causal Google trend queries[J]. Journal of Air Transport Management, 2021, 95: No. 102083. 10.1016/j.jairtraman.2021.102083 |
2 | DANTAS T M, CYRINO OLIVEIRA F L, REPOLHO H M V. Air transportation demand forecast through Bagging Holt Winters methods[J]. Journal of Air Transport Management, 2017, 59: 116- 123. 10.1016/j.jairtraman.2016.12.006 |
3 | 梁小珍, 乔晗, 汪寿阳, 等. 基于奇异谱分析的我国航空客运量集成预测模型[J]. 系统工程理论与实践, 2017, 37( 6): 1479- 1488. 10.12011/1000-6788(2017)06-1479-10 |
LIANG X Z, QIAO H, WANG S Y, et al. An integrated forecasting model for air passenger traffic in China based on singular spectrum analysis[J]. Systems Engineering — Theory and Practice, 2017, 37( 6): 1479- 1488. 10.12011/1000-6788(2017)06-1479-10 | |
4 | TSUI W H K, BALLI H O, GILBEY A, et al. Forecasting of Hong Kong Airport’s passenger throughput[J]. Tourism Management, 2014, 42: 62- 76. 10.1016/j.tourman.2013.10.008 |
5 | NIETO M R, CARMONA-BENÍTEZ R B. ARIMA + GARCH + Bootstrap forecasting method applied to the airline industry[J]. Journal of Air Transport Management, 2018, 71: 1- 8. 10.1016/j.jairtraman.2018.05.007 |
6 | GUNTER U, ZEKAN B. Forecasting air passenger numbers with a GVAR model[J]. Annals of Tourism Research, 2021, 89: No. 103252. 10.1016/j.annals.2021.103252 |
7 | NOURZADEH F, EBRAHIMNEJAD S, KHALILI-DAMGHANI K, et al. Forecasting the international air passengers of Iran using an artificial neural network[J]. International Journal of Industrial and Systems Engineering, 2020, 34( 4): 562- 581. 10.1504/ijise.2020.106089 |
8 | BAO Y K, XIONG T, HU Z Y. Forecasting air passenger traffic by support vector machines with ensemble empirical mode decomposition and slope-based method[J]. Discrete Dynamics in Nature and Society, 2012, 2012: No. 431512. 10.1155/2012/431512 |
9 | QIN L, LI W D, LI S J. Effective passenger flow forecasting using STL and ESN based on two improvement strategies[J]. Neurocomputing, 2019, 356: 244- 256. 10.1016/j.neucom.2019.04.061 |
10 | WAN H Y, GUO S N, YIN K, et al. CTS-LSTM: LSTM-based neural networks for correlated time series prediction[J]. Knowledge-Based Systems, 2020, 191: No. 105239. 10.1016/j.knosys.2019.105239 |
11 | 朱贾悦, 方雨晨, 李欣月. 成都双流机场旅客吞吐量的预测研究[J]. 交通科技与经济, 2019, 21( 1): 48- 51, 69. |
ZHU J Y, FANG Y C, LI X Y. Prediction research on passenger throughput of Chengdu Shuangliu Airport[J]. Technology and Economy in Areas of Communications, 2019, 21( 1): 48- 51, 69. | |
12 | ZHAO K Y, LI L, CAI S H, et al. An optimized time series combined forecasting method based on neural networks[J]. The Journal of Supercomputing, 2020, 76( 4): 2986- 3012. 10.1007/s11227-019-03064-5 |
13 | JIN F, LI Y W, SUN S L, et al. Forecasting air passenger demand with a new hybrid ensemble approach[J]. Journal of Air Transport Management, 2020, 83: No. 101744. 10.1016/j.jairtraman.2019.101744 |
14 | 梁小珍, 郭战坤, 张倩文, 等. 基于奇异谱分析的航空客运需求分析与分解集成预测模型[J]. 系统工程理论与实践, 2020, 40( 7): 1844- 1855. 10.12011/1000-6788-2019-1010-12 |
LIANG X Z, GUO Z K, ZHANG Q W, et al. An analysis and decomposition ensemble prediction model for air passenger demand based on singular spectrum analysis[J]. Systems Engineering — Theory and Practice, 2020, 40( 7): 1844- 1855. 10.12011/1000-6788-2019-1010-12 | |
15 | NAI W, LIU L, WANG S Y, et al. An EMD-SARIMA-based modeling approach for air traffic forecasting[J]. Algorithms, 2017, 10( 4): No. 139. 10.3390/a10040139 |
16 | 梁小珍, 邬志坤, 杨明歌, 等. 基于二次分解策略和模糊时间序列模型的航空客运需求预测研究[J]. 中国管理科学, 2020, 28( 12): 108- 117. |
LIANG X Z, WU Z K, YANG M G, et al. Air passenger demand forecasting based on dual decomposition strategy and fuzzy time series model[J]. Chinese Journal of Management Science, 2020, 28( 12): 108- 117. | |
17 | YANG S B, DENG Z G, LI X F, et al. A novel hybrid model based on STL decomposition and one-dimensional convolutional neural networks with positional encoding for significant wave height forecast[J]. Renewable Energy, 2021, 173: 531- 543. 10.1016/j.renene.2021.04.010 |
18 | 董永峰, 孙跃华, 高立超, 等. 基于改进一维卷积和双向长短期记忆神经网络的故障诊断方法[J]. 计算机应用, 2022, 42( 4): 1207- 1215. 10.11772/j.issn.1001-9081.2021071243 |
DONG Y F, SUN Y H, GAO L C, et al. Fault diagnosis method based on improved one-dimensional convolution and bidirectional long short-term memory neural networks[J]. Journal of Computer Applications, 2022, 42( 4): 1207- 1215. 10.11772/j.issn.1001-9081.2021071243 | |
19 | WU Q L, LIN H X. Daily urban air quality index forecasting based on variational mode decomposition, sample entropy and LSTM neural network[J]. Sustainable Cities and Society, 2019, 50: No. 101657. 10.1016/j.scs.2019.101657 |
20 | GUO Z, YU B, HAO M Y, et al. A novel hybrid method for flight departure delay prediction using Random Forest Regression and Maximal Information Coefficient[J]. Aerospace Science and Technology, 2021, 116: No. 106822. 10.1016/j.ast.2021.106822 |
21 | 李莎, 林晖. 结合MLR和ARIMA模型的时空建模及预测[J]. 计算机工程与应用, 2021, 57( 13): 276- 282. 10.3778/j.issn.1002-8331.2004-0214 |
LI S, LIN H. Spatio-temporal modelling and prediction combined with MLR and ARIMA model[J]. Computer Engineering and Applications, 2021, 57( 13): 276- 282. 10.3778/j.issn.1002-8331.2004-0214 | |
22 | XU S J, CHAN H K, ZHANG T T. Forecasting the demand of the aviation industry using hybrid time series SARIMA-SVR approach[J]. Transportation Research Part E: Logistics and Transportation Review, 2019, 122: 169- 180. 10.1016/j.tre.2018.12.005 |
23 | SUN S L, WEI Y J, TSUI K L, et al. Forecasting tourist arrivals with machine learning and internet search index[J]. Tourism Management, 2019, 70: 1- 10. 10.1016/j.tourman.2018.07.010 |
24 | DO Q H, LO S K, CHEN J F, et al. Forecasting air passenger demand: a comparison of LSTM and SARIMA[J]. Journal of Computer Science, 2020, 16( 7): 1063- 1084. 10.3844/jcssp.2020.1063.1084 |
25 | 龚飘怡, 罗云峰, 方哲梅, 等. 基于Attention-BiLSTM-LSTM神经网络的短期电力负荷预测方法[J]. 计算机应用, 2021, 41( S1): 81- 86. 10.11772/j.issn.1001-9081.2020091423 |
GONG P Y, LUO Y F, FANG Z M, et al. Short-term power load forecasting method based on Attention-BiLSTM-LSTM neural network[J]. Journal of Computer Applications, 2021, 41( S1): 81- 86. 10.11772/j.issn.1001-9081.2020091423 | |
26 | 熊涛. 基于EMD的时间序列预测混合建模技术及其应用研究[D]. 武汉: 华中科技大学, 2014: 50- 51. |
XIONG T. EMD-based hybrid modeling techniques for time series forecasting and their applications[D]. Wuhan: Huazhong University of Science and Technology, 2014: 50- 51. |
[1] | 黄云川, 江永全, 黄骏涛, 杨燕. 基于元图同构网络的分子毒性预测[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2964-2969. |
[2] | 潘烨新, 杨哲. 基于多级特征双向融合的小目标检测优化模型[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2871-2877. |
[3] | 李顺勇, 李师毅, 胥瑞, 赵兴旺. 基于自注意力融合的不完整多视图聚类算法[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2696-2703. |
[4] | 秦璟, 秦志光, 李发礼, 彭悦恒. 基于概率稀疏自注意力神经网络的重性抑郁疾患诊断[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2970-2974. |
[5] | 王熙源, 张战成, 徐少康, 张宝成, 罗晓清, 胡伏原. 面向手术导航3D/2D配准的无监督跨域迁移网络[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2911-2918. |
[6] | 刘禹含, 吉根林, 张红苹. 基于骨架图与混合注意力的视频行人异常检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2551-2557. |
[7] | 顾焰杰, 张英俊, 刘晓倩, 周围, 孙威. 基于时空多图融合的交通流量预测[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2618-2625. |
[8] | 石乾宏, 杨燕, 江永全, 欧阳小草, 范武波, 陈强, 姜涛, 李媛. 面向空气质量预测的多粒度突变拟合网络[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2643-2650. |
[9] | 赵亦群, 张志禹, 董雪. 基于密集残差物理信息神经网络的各向异性旅行时计算方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2310-2318. |
[10] | 徐松, 张文博, 王一帆. 基于时空信息的轻量视频显著性目标检测网络[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2192-2199. |
[11] | 孙逊, 冯睿锋, 陈彦如. 基于深度与实例分割融合的单目3D目标检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2208-2215. |
[12] | 吴筝, 程志友, 汪真天, 汪传建, 王胜, 许辉. 基于深度学习的患者麻醉复苏过程中的头部运动幅度分类方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2258-2263. |
[13] | 李欢欢, 黄添强, 丁雪梅, 罗海峰, 黄丽清. 基于多尺度时空图卷积网络的交通出行需求预测[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2065-2072. |
[14] | 张郅, 李欣, 叶乃夫, 胡凯茜. 基于暗知识保护的模型窃取防御技术DKP[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2080-2086. |
[15] | 赵雅娟, 孟繁军, 徐行健. 在线教育学习者知识追踪综述[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1683-1698. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||