基于组合预测方法的天津港物流需求预测研究
Research on forecasting of Tianjin Port logistics demand based on combined prediction methods
叶璐瑶1
作者信息
- 1. 武汉科技大学管理学院，武汉，430065
- 折叠
摘要
天津港作为我国北方重要的综基于组合预测方法的天津港物流需求预测研究基于组合预测方法的天津港物流需求预测研究合性港口，其货物吞吐量水平直接反映区域经济发展状况及港口运营能力。对天津港货物吞吐量进行科学预测，对于港口资源配置、运输规划及政策制定具有重要现实意义。首先，通过对天津港货物吞吐量2005-2025年季度数据进行趋势与波动特征分析，发现该序列具有明显的上升趋势和阶段性波动特征。在此基础上构建滚动GM(k,1)模型以刻画长期趋势，并利用BP神经网络对残差进行非线性修正，建立BP-GM残差修正预测模型，随后引入Shapley值方法确定模型权重，构建组合预测模型，实现多模型信息融合。通过对2025-2027年天津港货物吞吐量进行预测，结果表明，与单一滚动GM(k,1)模型相比，BP-GM残差修正模型与Shapley值组合模型在预测精度方面均表现出明显优势，预测误差分别降低45.7%和41.2%，为港口吞吐量预测提供了一种可行的技术路径。
Abstract
Tianjin Port, as an important comprehensive port in northern China, has a cargo throughput level that directly reflects regional economic development and port operational capacity. Scientifically forecasting the cargo throughput of Tianjin Port is of great practical significance for port resource allocation, transportation planning, and policy formulation.First, by analyzing the trend and fluctuation characteristics of the quarterly cargo throughput data of Tianjin Port from 2005 to 2025, it is found that the series exhibits a clear upward trend and stage-wise fluctuation patterns. On this basis, a rolling GM(k,1) model is constructed to capture the long-term trend, and a BP neural network is employed to perform nonlinear correction on the residuals, thereby establishing a BP-GM residual correction forecastiResearch on forecasting of Tianjin Port logistics demand based on combined prediction methodsng model. Subsequently, the Shapley value method is introdResearch on forecasting of Tianjin Port logistics demand based on combined prediction methodsuced to determine model weights and construct a combined forecasting model, enabling the integration of information from multiple models.By forecasting the cargo throughput of Tianjin Port for 2025-2027, the results indicate that, compared with the single rolling GM(k,1) model, both the BP-GM residual correction model and the Shapley value-based combined model demonstrate significant improvements in forecasting accuracy, with prediction errors reduced by 45.7% and 41.2%, respectively. The proposed approach provides a feasible technical pathway for port cargo throughput forecasting.关键词
天津港货物吞吐量/灰色预测模型/BP神经网络/夏普利值/残差修正Key words
Cargo throughput of Tianjin Port/gray prediction model/BP neutal network/shapley value/residual correction引用本文复制引用
叶璐瑶.基于组合预测方法的天津港物流需求预测研究[EB/OL].(2026-03-19)[2026-03-21].http://www.paper.edu.cn/releasepaper/content/202603-182.学科分类
交通运输经济/综合运输
评论