基于LightGBM模型的产品订单需求量预测  

Product Order Demand Forecast Based on LightGBM Model

作　　者：赵玉骁 陈思予 叶凯圳 施锋伟 金秀玲[1] 

机构地区：[1]闽江学院数学与数据科学学院,福建 福州

出　　处：《应用数学进展》2023年第11期4708-4716,共9页Advances in Applied Mathematics

摘　　要：产品订单需求量预测是管理企业供应链的关键环节。准确预测客户对产品的需求量是很有必要的。为了解决不同产品的需求量问题,本文充分利用厂商数据,采用基于LightGBM的集成算法建立产品订单量预测模型。用网格探索进行参数调优,用3折目标编码,最终测试集上的MAPE为0.3541%;拟合效果良好且泛化能力强。最后用LightGBM模型预测后三个月的各个地区、各个品类的月度订单需求量。有助于企业资源有效配置,提高企业的收益效率,具有较大的现实意义和参考价值。Product order demand forecasting is a key step in managing enterprise supply chain. It is necessary to accurately predict the customer’s demand for the product. In order to address the demand fore-casting challenge for different products, this article leverages vendor data extensively and employs an ensemble algorithm based on LightGBM to build a predictive model for product order quantities. Mesh exploration was used for parameter tuning and 3-fold target coding. The MAPE on the final test set was 0.3541%. The fitting effect is good and the generalization ability is strong. Finally, LightGBM model is used to forecast the monthly order demand of each region and category in the next three months. It is helpful to the effective allocation of enterprise resources and improve the profit efficiency of enterprises, and has great practical significance and reference value.

关 键 词：订单需求预测 机器学习 LightGBM 

分 类 号：F42[经济管理—产业经济]