[1]刘雁兵,孔维力,刘晓蓉,等.基于多模态信息融合的卷烟销量预测方法[J].江西师范大学学报(自然科学版),2023,(05):497-505.[doi:10.16357/j.cnki.issn1000-5862.2023.05.09]
　LIU Yanbing,KONG Weili,LIU Xiaorong,et al.The Multimodal-Based Method for Tobacco Sales Forecast[J].Journal of Jiangxi Normal University:Natural Science Edition,2023,(05):497-505.[doi:10.16357/j.cnki.issn1000-5862.2023.05.09]
点击复制

基于多模态信息融合的卷烟销量预测方法()
分享到：

《江西师范大学学报》（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2023年05期
页码:
497-505
栏目:
信息科学与技术
出版日期:
2023-09-25

文章信息/Info

Title:
The Multimodal-Based Method for Tobacco Sales Forecast
文章编号:
1000-5862(2023)05-0497-09
作者:
刘雁兵1; 孔维力1;  刘晓蓉2;  王义新2;  汪伟飞3
(1.广西中烟工业有限责任公司,广西 南宁 530001; 2.广东烟草广州市有限公司,广东 广州 510510; 3.武汉人工智能研究院,湖北 武汉 430010)
Author(s):
LIU Yanbing1; KONG Weili1;  LIU Xiaorong2;  WANG Yixin2;  WANG Weifei3
(1.China Tobacco Guangxi Industrial Company Limited,Nanning Guangxi 530001, China; 2.Guangdong Tobacco Guangzhou Company Limited, Guangzhou Guangdong 510510, China; 3. Wuhan AI Research,Wuhan Hubei 430010, China)
关键词:
深度学习;  图像生成;  目标检测;  卷烟识别;  BERT模型
Keywords:
deep learning;  image generation;  object detection;  tobacco recognition;  BERT
分类号:
TP 391.4
DOI:
10.16357/j.cnki.issn1000-5862.2023.05.09
文献标志码:
A
摘要:
融合多模态信息的数据科学对智能营销至关重要.该文提出了一种融合了视觉、自然语言和结构化数据的基于多模态信息的烟草销售预测方法.首先,引入扩散模型来生成高质量的香烟图像样本以进行数据扩增,在柜台香烟识别阶段,采用深度耦合网络和多元组排序损失来提高香烟识别率; 其次,在销售预测中,使用多模态信息作为输入,包括柜台位置、采用双向文本编码的品牌表示以及相应价格; 最后,通过预测算法得到香烟的预测销量.通过全面综合的分析为营销提供了有价值的建议,促进了多模态信息在卷烟营销科学上的应用.
Abstract:
Data science with multimodal information is of great importance for intelligent marketing. In the paper, a tobacco sales prediction method based on multimodal information is proposed, including visual, natural language, and structured data. Firstly, a diffusion model is introduced to generate high-quality cigarette image samples for cigarette recognition. In the cigarette recognition stage, a deep coupled network and multiple sets of ranking losses are employed to improve cigarette recognition at the counter.Secondly,in sales prediction,cigarettes are represented with multimodal information,including counter location,brand representation encoded with Bidirectional Encoders Representations from Transformers(BERT),and corresponding prices.Finally,the monthly sales of cigarettes are provided. Through comprehensive analysis, valuable strategic recommendations are provided for marketing, promoting the multimodal-based application in tobacco science marketing.

参考文献/References:

[1] KRAMOLISˇ J, KOPECˇKOVA' M. Product placement: a smart marketing tool shifting a company to the next competitive level [J].Journal of Competitiveness,2013,5(4):98-114.
[2] DE ANDRADE M, HASTINGS G, ANGUS K.Promotion of electronic cigarettes: tobacco marketing reinvented [EB/OL].[2022-09-13].https://doi.org/10.1136/bmj.f7473.
[3] SIMO··ES D M,BARBOSA B,FILIPE S.Smart marketing with the internet of things [EB/OL].[2022-09-19]. https://www.igi-global.com/gateway/book/192038.
[4] DHAR V.Data science and prediction [J].Communica-tions of the ACM,2013,56(12):64-73.
[5] 蒋庆,陶峰,章玉,等.基于大数据分析的车辆安全检测系统研究 [J].江西师范大学学报(自然科学版),2022,46(6):649-654.
[6] 苗永春,程艳.离群点检测方法及其在大数据时代下的改进方法研究 [J].江西师范大学学报(自然科学版),2014,38(5):454-458,495.
[7] 穆建军.中小型烟草零售终端卷烟陈列方式初探 [J].科技风,2012(16):192.
[8] 刘薇.卷烟陈列标准研究[J].商,2014(25):244-245,241.
[9] 何金宝,胡秋宝,付志超,等.基于DCGAN和改进YOLOv5s的桥梁表面缺陷检测识别 [J].江西师范大学学报(自然科学版),2022,46(6):655-660.
[10] 肖雷雨,王澍,刘渊根,等.基于深度学习技术的烟梗形态分类与识别 [J].烟草科技,2021,54(6):65-74.
[11] 梁冬,陈智斌,农英雄,等.基于图像识别技术的卷烟零售数据采集方法 [J].中国烟草学报,2021,27(5):81-89.
[12] 单宇翔,龙涛,楼卫东,等.基于深度学习的复杂场景中卷烟烟盒检测与识别方法 [J].中国烟草学报,2021,27(5):71-80.
[13] ZHANG Han, XU Tao, LI Hongsheng, et al.StackGAN: text to photo-realistic image synthesis with stacked generative adversarial networks [EB/OL].[2022-09-26].https://ieeexplore.ieee.org/document/8411144.
[14] XU Tao,ZHANG Pengchuan,HUANG Qiuyuan,et al. AttnGAN:Fine-grained text to image generation with attentional generative adversarial networks [EB/OL]. [2022-09-28].https://arxiv.org/pdf/1711.10485.pdf.
[15] DHARIWAL P,NICHOL A. Diffusion models beat gans on image synthesis [EB/OL].[2022-09-23]. https://arxiv.org/abs/2105.05233.
[16] JONATHAN H,SALIMANS T.Classifier-free diffusion guidance [EB/OL].[2022-09-12]. https://blog.csdn.net/NGUever15/article/details/128282683.
[17] GARDNER E S. Exponential smoothing:the state of the art [J].Journal of Forecasting,2006,22(4):637-666.
[18] 赵亚雄,唐立新.基于粗糙集多元回归的湖北物流需求预测 [J].现代商贸工业,2019,40(2):31-32.
[19] 谢乃明, 刘思峰, 离散GM(1,1)模型与灰色预测模型建模机理 [J].系统工程理论与实践,2005,25(1):93-99.
[20] 孙晓静,高慧,陈云.基于聚类分析和决策树算法的服装销售预测模型 [J].中国管理信息化,2015,18(9):64-67.
[21] 胡瑛,罗银,张瀚文,等.基于注意力机制的3D卷积神经网络孤立词手语识别 [J].湖南工程学院学报(自然科学报),2022,32(1):55-60.
[22] 吴娇,王鹏,乔梦雨,等.基于改进多级特征金字塔的陆战场目标检测算法 [J].计算机应用与软件,2020,37(10):155-161.

备注/Memo

备注/Memo:
收稿日期:2023-02-11
基金项目:国家自然科学基金(62076235)资助项目.
作者简介:刘雁兵(1978—),男,湖北洪湖人,博士研究生,主要从事计算机视觉和大数据技术应用的研究.E-mail:wangweiwei@wair.ac.cn
更新日期/Last Update: 2023-09-25