基于GA-SVR的江苏省区域物流需求预测研究  

Optimization GA-SVR Logistics Demand of Jiangsu Province Prediction Model on Genetic Algorithm

作　　者：王娜 李晏丞[1] WANG Na;LI Yancheng(Logistics Management College of Wuxi Transportation Branch of Jiangsu United Vocational and Technical College,Wuxi 214151,China)

机构地区：[1]江苏联合职业技术学院无锡交通分院物流管理学院,江苏无锡214151

出　　处：《物流科技》2025年第5期101-105,共5页Logistics Sci Tech

基　　金：中国交通教育研究会教育科学研究课题项目(JT2022ZD069、JT2022YB288)。

摘　　要：现代物流产业作为我国经济发展的支柱产业之一,物流产业发展可以提高劳动的生产效率降低资源消耗,是经济高速发展的基础。为了科学合理制定物流产业发展规划,需要准确预测市场对物流的需求量,根据物流需求情况来制定物流发展目标,才能更好满足经济发展的需求。文章对江苏省1990—2021年经济发展数据进行研究分析,借助于机器学习方法,构建了经济指标与物流需求之间关系的支持向量回归(SVR)黑箱模型。针对SVR核函数参数难以确定的问题,采用遗传算法对其寻优,并与DE-SVR、GS-SVR和LS等方法对比货运量回测结果,基于遗传算法优化的SVR黑箱模型(GA-SVR)具有精度高、训练速度快的优点,更适用于当前物流需求数据集。通过GA-SVR建立了货运量时间序列的预测模型,对2019—2021年货运量进行预测,预测结果与真实值之间最大误差仅为3.45%,证明了所训练时间序列模型的有效性和泛化性能,并采用预测模型预测了江苏省未来三年货运量。The modern logistics industry is one of the pillar industries in China's economic development.The development of the logistics industry can improve labor productivity and reduce resource consumption,serving as the foundation for the rapid economic development.To formulate a scientific and reasonable logistics industry development plan,it is necessary to accurately predict the market demand for logistics and set logistics development goals based on the logistics demand situation,so as to better meet the requirements of economic development.By conducting research and analysis on the economic development data of Jiangsu Province from 1990 to 2021 and with the aid of machine learning methods,a support vector regression(SVR)black-box model of the relationship between economic indicators and logistics demand was constructed.In view of the difficulty in determining the parameters of the SVR kernel function,the genetic algorithm was adopted to optimize them.And by comparing the backtesting results of freight volume with methods such as DE-SVR,GS-SVR and LS,the SVR black-box model optimized by the genetic algorithm(GA-SVR)has the advantages of high precision and fast training speed,and is more suitable for the current logistics demand data set.Through GA-SVR,a prediction model for the time series of freight volume was established to predict the freight volume from 2019 to 2021.The maximum error between the predicted results and the actual values was only 3.45%,which proved the effectiveness and generalization performance of the trained time series model.Moreover,the prediction model was used to predict the freight volume of Jiangsu Province in the next three years.

关 键 词：物流需求 数学模型 指标体系 预测对比 

分 类 号：F259.27[经济管理—国民经济]