基于GM-LSTM的港口物流需求预测——以宁波港域为例  

Port Logistics Demand Forecasting Based on GM-LSTM——A Case Study of Ningbo Port Area

作　　者：江帆 刘利民[1] JIANG Fan;LIU Limin(Zhejiang Wanli University,Ningbo 315121,China)

机构地区：[1]浙江万里学院,浙江宁波315121

出　　处：《物流科技》2024年第24期25-28,50,共5页Logistics Sci Tech

基　　金：浙江省哲学社会科学后期资助项目(23HQZZ33YB)。

摘　　要：文章针对现有港口物流需求预测的传统统计方法和机器学习技术,如ARMA模型等方法在处理复杂、非线性和长短期依赖特性数据时的不足,提出了一种结合GM (1,1)和LSTM的组合预测模型。对比ARMA法、GM (1,1)、BP神经网络和LSTM网络的预测效果,仿真结果表明:GM-LSTM组合模型在准确性和稳健性方面有显著提升。该组合模型能够有效预测宁波港域2023—2027年的物流需求,为港口管理决策提供了科学依据。研究成果有助于优化宁波港域的资源配置,降低物流成本,在国际竞争中保持其竞争优势,且为同类型港口提供了一种高效的物流需求预测方法。The traditional statistical methods and machine-learning techniques such as ARMA model face limitations when dealing with complex,nonlinear,and long-short term dependent data in predicting existing port logistics demands.In response,a combined forecasting model integrating GM(1,1)and LSTM is proposed.Comparative analysis against ARMA,GM(1,1),BP neural network,and LSTM models demonstrates significant improvements in accuracy and robustness with the GM-LSTM combination.Simulation results indicate the effectiveness of this combined model in accurately predicting the logistics demand at Ningbo Port from 2023 to 2027,providing a scientific basis for port management decisions.The research findings contribute to optimizing resource allocation at Ningbo Port,reducing logistics costs,and maintaining its competitive edge in the international arena.Moreover,it offers an efficient logistics demand prediction method applicable to similar ports worldwide.

关 键 词：港口物流需求预测 GM(1 1) 长短期记忆网络 组合预测 

分 类 号：F201[经济管理—国民经济]