基于TEI@I的港口集装箱吞吐量预测方法研究  

Forecasting of Port Container Throughput Based on TEI@I Methodology

作　　者：贾鹏 陆圣斓 邬桐 汪寿阳 JIA Peng;LU Shenglan;WU Tong;WANG Shouyang(Collaborative Innovation Center for Transport Studies,Dalian Maritime University,Dalian 116026;School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100190;Academy of Mathematics and Systems Sciences,Chinese Academy of Sciences,Beijing 100190)

机构地区：[1]大连海事大学综合交通运输协同创新中心,大连116026 [2]中国科学院大学经济与管理学院,北京100190 [3]中国科学院数学与系统科学研究院,北京100190

出　　处：《系统科学与数学》2022年第12期3321-3338,共18页Journal of Systems Science and Mathematical Sciences

基　　金：国家自然科学基金项目(72174035);辽宁省“兴辽英才计划”项目(XLYC2008030);高等学校学科创新引智计划(B20082);中国博士后科学基金会面上项目(2015M580128);中央高校基本科研业务费专项资金(3132022641)资助课题。

摘　　要：文章基于TEI@I (Text mining+Econometrics+Intelligent algorithms@Integration)方法论,提出了一种能够适应外部突发事件的港口集装箱吞吐量预测框架.首先对港口集装箱吞吐量原始序列进行互补集合经验模态分解(CEEMD),得到不同时间尺度的子序列,并分析各子序列的复杂性,平稳性和(与原序列的)相关性,在此基础上分类重构新的子序列;然后根据各子序列的不同特征构建和选择适配的预测模型,对于外部突发事件造成的冲击部分,引入改进的标准事件分析法量化不规则大事件的影响方向,影响程度和影响时长;最后集成各部分的预测结果并结合“专家系统”进行优化得到最终预测结果.通过上海港及青岛港的实证研究,表明该方法能够在发生重大不规则事件的情况下保证其预测精度,预测方向和模型稳定性,其预测结果优于其他基准模型.Based on the TEI@I(Text mining+Econometrics+Intelligent algorithms@Integration) methodology,this paper proposes a port container throughput forecasting framework that can adapt to external contingencies.Firstly,the original sequence of port container throughput is decomposed by CEEMD(Complementary Ensemble Empirical Modal Decomposition) to obtain subseries of different time scales,and each subseries is analyzed for complexity,smoothness and correlation(with the original sequence),on which the new subseries are classified and reconstructed;then adapted forecasting models are constructed and selected according to the different characteristics of each subseries;For the part of huge shocks caused by external contingencies,an improved standard event analysis method is introduced to quantify the impact direction,impact degree and impact duration of irregular large events;finally,the prediction results of each part are integrated and combined with the "expert system" to optimize the final prediction results.The empirical studies of Shanghai and Qingdao ports show that the method is able to guarantee its prediction accuracy,prediction direction and model stability in the presence of major irregularities,and the results are better than other benchmark models.

关 键 词：水路运输 TEI@I方法论 KNN-LSTM 集装箱吞吐量 集成预测模型 

分 类 号：U652.14[交通运输工程—港口、海岸及近海工程]