%0 Journal Article %T 基于LSTM的杭州市智能物流需求预测与优化研究
Research on Demand Forecasting and Optimization of Hangzhou Intelligent Logistics Based on LSTM %A 信嘉兴 %J E-Commerce Letters %P 3673-3683 %@ 2168-5851 %D 2025 %I Hans Publishing %R 10.12677/ecl.2025.1451685 %X 随着经济的持续发展和国际贸易的快速扩张，我国物流产业呈现出迅猛的增长态势。然而，在实际操作中，许多决策者却低估了物流需求预测的重要性，这种忽视往往导致资源的极大浪费和社会成本的增加。因此，构建科学且实用的物流需求预测模型，并实现精准预测，已成为推动我国物流行业持续健康发展的重要基础和关键前提。本研究设计了一种基于注意力机制的双层LSTM架构，增强了模型捕捉长短期依赖关系的能力，此外还构建了适合杭州市电商物流特点的特征工程方法，引入电商交易指数、季节性指标等本地化变量，在此基础上提出了MSLE-MAE组合损失函数和模型集成学习框架，有效提高了预测精度，实验结果表明，该模型在短期预测中准确率达95.2%，具有显著的理论和实践价值。
With the sustained development of economy and the rapid expansion of international trade, our country’s logistics industry has shown rapid growth. However, in practice, many decision makers underestimate the importance of logistics demand forecasting, which often leads to a great waste of resources and an increase in social costs. Therefore, the construction of a scientific and practical logistics demand forecasting model and the realization of accurate forecasting has become an important foundation and key premise to promote the sustainable and healthy development of China’s logistics industry. In this study, a two-layer LSTM architecture based on attention mechanism was designed to enhance the ability of the model to capture long-term and short-term dependencies. In addition, a feature engineering method suitable for the characteristics of Hangzhou e-commerce logistics was constructed, and localized variables such as e-commerce transaction index and seasonal index were introduced. On this basis, the MSLE-MAE combined loss function and model integrated learning framework are proposed, which effectively improves the prediction accuracy. The experimental results show that the accuracy of the model in short-term prediction is 95.2%, which has significant theoretical and practical value. %K LSTM， %K 物流需求， %K 数字经济， %K 智能化
LSTM %K Logistics Needs %K Digital Economy %K Intelligentize %U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=116403