中文核心期刊
CSCD来源期刊
中国科技核心期刊
RCCSE中国核心学术期刊
重庆交通大学学报（自然科学版）

重庆交通大学学报（自然科学版） ›› 2024, Vol. 43 ›› Issue (3): 92-98.DOI: 10.3969/j.issn.1674-0696.2024.03.11

• 交通+大数据人工智能 • 上一篇    

基于AOA-LSSVM模型的枢纽城市物流需求量预测

肖红1，2，夏如玉1，2，王孝坤3，4，杨雪峰5   

  1. (1. 重庆交通大学 经济与管理学院，重庆 400074； 2. 重庆口岸物流与航运发展研究中心，重庆 400074； 3. 大连交通大学 经济管理学院，辽宁 大连 116000； 4. 大连交通大学“一带一路”交通互联互通与人文交流研究院， 辽宁 大连 116003; 5. 辽宁省交通运输事业发展中心，辽宁 沈阳 110005)
  • 收稿日期:2023-03-02 修回日期:2023-12-19 发布日期:2024-03-21
  • 作者简介:肖 红（1969—），女，四川江安人，教授，博士，主要从事交通运输方面的研究。E-mail:158564498@qq.com 通信作者：王孝坤（1975—），男，辽宁大连人，教授，博士，主要从事物流规划方面的研究。E-mail:24014902@qq.com
  • 基金资助:
    国家社科基金一般项目（21BJY223）；国家自然科学基金项目（71864022）；重庆市教育委员会人文社会科学基金（21SKGH088）

Prediction of Logistics Demand in Hub Cities Based on AOA-LSSVM Model

XIAO Hong1，2, XIA Ruyu1，2, WANG Xiaokun3，4, YANG Xuefeng5   

  1. (1. School of Economics and Management, Chongqing Jiaotong University, Chongqing 400074, China; 2. Chongqing Port Logistics and Shipping Development Research Centre, Chongqing 400064, China; 3. School of Economics and Management, Dalian Jiaotong University, Dalian 116000, Liaoning, China; 4. “The Belt and Road” Traffic Interoperability and Humanities Exchange Research Institute, Dalian Jiaotong University, Dalian 116003, Liaoning, China; 5. Transport Development Centre of Liaoning Province, Shenyang 110005, Liaoning, China)
  • Received:2023-03-02 Revised:2023-12-19 Published:2024-03-21

PDF

摘要： 传统的LSSVM难以全面反映物流需求的变化规律，会导致预测效果不佳。首先利用灰色关联分析（GRA）得到物流需求的主要影响因素；将主要影响因素作为LSSVM的输入变量，构建物流需求预测模型；通过阿基米德算法（AOA）对最小二乘支持向量机的正则化参数（γ）和核参数（σ）进行迭代寻优，以减少参数选择的盲目性；构建AOA算法优化最小二乘支持向量机（LSSVM）的智能预测模型AOA-LSSVM，经过验证该模型可以提高预测精度。运用AOA-LSSVM模型对西部陆海新通道的重要枢纽城市——重庆、成都、贵阳和南宁的物流需求进行实证分析，结果表明：该模型与LSSVM模型相比取得较高的预测精度，其均方根误差、平均绝对误差、以及异方差调整的均方根误差、异方差调整的平均绝对误差分别降低了1 946.4，1 206.1，0.028 4，0.039 7。

关键词: 交通运输工程；AOA算法；LSSVM模型；西部陆海新通道；物流需求预测

Abstract: Traditional LSSVM (least squares support vector machine) is difficult to comprehensively reflect the change pattern of logistics demand, which will cause poor prediction performance. Grey correlation analysis (GRA) was used to obtain the main influencing factors of logistics demand, and then the main influencing factors were used as input variables of LSSVM to construct a logistics demand forecasting model. The regularization parameters (γ) and kernel parameters (σ) of LSSVM were iteratively optimized by the Archimedes algorithm (AOA) to reduce the blindness of parameter selection. Thus, an intelligent prediction model AOA-LSSVM was constructed to optimize LSSVM by using the AOA algorithm. After verification, the proposed model can improve prediction accuracy. The proposed AOA-LSSVM model was used to empirically analyze the logistics demand of the important hub cities of the western land and sea new passage such as Chongqing, Chengdu, Guiyang and Nanning. The results show that compared to the LSSVM model, the proposed model can achieve higher prediction accuracy, with a reduction of 1 946.4, 1 206.1, 0.028 4 and 0.039 7 in root mean square error, mean absolute error, and heteroskedasticity-adjusted root mean square error and mean absolute error, respectively.

Key words: traffic and transportation engineering; AOA algorithm; LSSVM model; new western land-sea passage; logistics demand forecast

中图分类号: