基于LSTM-LightGBM模型的烟草存销比层次化预测方法
A Hierarchical Forecasting Method for Tobacco Inventory-to-Sales Ratio Based on LSTM-LightGBM Model
李家蕊 1杨旻1
作者信息
- 1. 烟台大学数学与信息科学学院,山东烟台 264005
- 折叠
摘要
基于LSTM和LightGBM算法,结合零售商的地理位置和档位信息,构建了 一个层次化模型,用于准确预测烟草产品的存销比.该模型首先利用LSTM网络对不同地区和档位的整体存销比进行预测.随后,将得到的整体存销比作为LightGBM的辅助输入,用于预测个体零售商销售的每一类卷烟的存销比.本模型逐步融合数据的宏观和微观特征,对某地烟草实际销售数据的实验结果表明,LSTM-LightGBM模型具有较好的预测精度.
Abstract
The paper employs the LSTM-LightGBM algorithm and incorporates the geographical location and grade information of retailers to develop a hierarchical model for accurately forecasting the inventory-to-sales ratio of tobac-co products.The model initially uses the LSTM network to forecast the overall inventory-to-sales ratio across differ-ent regions and grades.Subsequently,the obtained overall inventory-to-sales ratio is utilized as supplementary in-put for LightGBM to predict the inventory-to-sales ratio for each type of cigarette sold by individual retailers.The proposed model progressively combines macro-and micro-level features of the data.The validation results,using actual tobacco sales data from a specific region,demonstrate the superior predictive accuracy of the proposed ap-proach.
关键词
烟草/存销比/LSTM/LightGBM/层次化模型Key words
tobacco/inventory-to-sales ratio/LSTM/LightGBM/hierarchical model引用本文复制引用
基金项目
山东省自然科学基金资助项目(ZR2021MA010)
出版年
2024
国家科技期刊平台
NSTL
万方数据