基于跨序列组合模型的饮料业需求预测  

Demand Forecast of Beverage Industry Based on Cross-series Combination Model

作　　者：刘松诺 陈敬贤 杨惠 夏焱 LIU Songnuo;CHEN Jingxian;YANG Hui;XIA Yan(School of Management,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]合肥工业大学管理学院,安徽合肥230009

出　　处：《物流科技》2024年第17期21-26,共6页Logistics Sci Tech

摘　　要：面对由多种现实因素引发的需求波动,饮料行业急需提升供应链效率,而精准的需求预测在这一过程中扮演着至关重要的角色。为解决这一问题,文章提出了一种先进的需求预测组合模型——跨序列模型。该模型通过“借用”其他产品线的销售数据,再输入多种预测模型。文章通过标准化平均绝对误差(NMAE)和标准化均方误差(NMSE)等性能评估指标,将其与传统的预测模型进行了详尽的对比分析。实证研究表明,该跨序列模型在“借用”其他产品销售数据的基础上,实现了比传统模型更高的预测精度,同时也产生了更为稳健和更符合逻辑的预测结果。Faced with demand fluctuations triggered by a variety of real-world factors,the beverage industry is in urgent need of enhancing supply chain efficiency,and accurate demand forecasting plays a pivotal role in this process.To address this issue,this study introduces an advanced demand forecasting composite model—a cross-series model.This model"borrows"sales data from other product lines and feeds it into various forecasting models.Through performance evaluation metrics such as Normalized Mean Absolute Error NMAE and Normalized Mean Square Error NMSE,this study provides a comprehensive comparative analysis with traditional forecasting models.Empirical research shows that,based on"borrowing"sales data from other products,the cross-series model achieves higher forecasting accuracy than traditional models,while also producing more robust and logical forecasting results.

关 键 词：需求预测 机器学习 跨序列训练 组合预测模型 

分 类 号：F272[经济管理—企业管理]