管理评论

›› 2018, Vol. 30 ›› Issue (1): 195-201.

• 物流与供应链管理 • 上一篇    下一篇

基于情境变动的港口吞吐量预测模型

鲁渤1,2, 杨显飞3, 汪寿阳4   

  1. 1. 大连大学国际学院, 大连 116622;
    2. 大连大学智慧航运与物流网络技术国家地方联合工程实验室, 大连 116622;
    3. 大连大学经济与管理学院, 大连 116622;
    4. 中国科学院数学与系统科学研究院, 北京 100190
  • 收稿日期:2017-02-16 出版日期:2018-01-28 发布日期:2018-01-24
  • 通讯作者: 杨显飞(通信作者),大连大学经济与管理学院讲师,博士。
  • 作者简介:鲁渤,大连大学国际学院副教授,硕士生导师,博士;汪寿阳,中国科学院数学与系统科学研究院教授,博士生导师,博士。
  • 基金资助:

    国家自然科学基金项目（71573028；71703011；71781330123）；辽宁高校杰出青年学者计划（WJQ2015004）；辽宁自然科学基金面上项目（201601006；20170540030）；大连市科技之星项目（2016RQ074）。

Port Throughput Forecasting Model Based on Context Change

Lu Bo1,2, Yang Xianfei3, Wang Shouyang4   

  1. 1. International College, Dalian University, Dalian 116622;
    2. National Joint Engineering Laboratory for Intelligent Shipping and Logistics Network Technology, Dalian University, Dalian 116622;
    3. School of Economics and Management, Dalian University, Dalian 116622;
    4. Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190
  • Received:2017-02-16 Online:2018-01-28 Published:2018-01-24

PDF (PC)

633

摘要：

情境变动是影响港口吞吐量预测的重要因素。然而现有关于港口吞吐量预测建模的研究对此问题的关注度相对较少。因此本文构建动态惩罚支持向量回归模型，通过动态调整每个数据的惩罚系数，使预测模型快速适应新情境下的事物发展规律，从而提高预测模型的预测准确率。采用1980-2014年大连港和天津港的年货物吞吐量数据进行实证研究，并与传统支持向量回归模型和ARIMA模型进行对比分析，实证结果表明本预测模型与其他两个预测模型相比具有以下两个特点：1、当影响港口吞吐量的情境发生变化时，该模型能够快速适应新情境，从而提高预测准确率；2、该模型的预测性能更加准确、平稳，从而提高实用性。

关键词: 情境变动, 港口吞吐量, 时间序列预测, 动态惩罚, 支持向量回归模型

Abstract:

Context change is an important factor that affects the forecasting of port throughput. However, existing researches on the modeling of port throughput forecasting are relatively less concerned about this issue. A dynamic penalty support vector regression model is proposed in this paper. By dynamically adjusting the penalty coefficient of each data, this model is quickly adapted to the development of things under the new contexts and improves prediction accuracy. An empirical research is carried out based on annual cargo throughput data of Dalian Port and Tianjin port in the period from 1980 to 2014, and a comparison is made with traditional support vector regression model and ARIMA model. Experimental results show that this model has the following two characteristics compared with the other two regression models:first, when the context change, the model can quickly adapt to new context, so as to improve the accuracy of prediction; second, the prediction performance of the model is more accurate and stable, so it is able to improve the practicality.

Key words: context change, port throughput, time series prediction, dynamic penalty, support vector regression model