应用机器学习对外卖含糖饮料消费的研究  

A Consumption Analysis of Sugar-sweetened Beverages Ordered Online for Home Delivery:An Application of Machine Learning

作　　者：翟倩倩 蒋玉 张家铭 胡安·埃斯特班·格瓦拉 王红[3] ZHAI Qianqian;JIANG Yu;ZHANG Jiaming;Juan Esteban GUEVARA;WANG Hong

机构地区：[1]浙江大学中国农村发展研究院,杭州310058 [2]西南财经大学中国西部经济研究院,成都611130 [3]普渡大学农业经济系,西拉法叶47907 [4]普渡大学统计系,西拉法叶47907

出　　处：《农业经济问题》2024年第7期20-33,共14页Issues in Agricultural Economy

基　　金：教育部人文社科重点研究基地重大项目“多重冲击与韧性食物系统研究”(编号:22JJD790079);四川省社会科学基金青年项目“数字经济背景下四川省粮食品牌发展路径与政策保障研究”(编号:SCJJ23ND418);教育部人文社会科学研究一般项目“食品公众政策与健康食品消费:基于含糖饮料消费的实证研究”(编号:21YJA790013)。

摘　　要：含糖饮料因其高热量而导致慢性疾病,引起的健康问题得到全球的关注。在外卖平台迅速发展的大背景下,本文基于中国六个城市的外卖平台的销售数据,运用机器学习方法识别外卖含糖饮料消费动态,并对价格折扣、糖税等模拟情景进行了销售变动分析。研究发现,第一,有监督机器学习模型中的神经网络相较于传统计量回归模型更加适用于消费趋势预测分析。第二,模型预测结果表明,当价格降低10%时,总销售额将比原始销售增加0.49%,说明含糖饮料消费富有价格弹性;此外,当价格上升10%时,总销量额将比原始销售下降0.17%,表明通过税收手段来减少含糖饮料的消费是可行策略。第三,外卖含糖饮料的销售情况存在时空差异,长三角地区外卖含糖饮料的销售量高于其他地区,周中的销售量高于周末时段,下午的销售量是一天的高峰。本文研究对于建立降低含糖饮料的过度消费的路径与机制,引导消费者健康饮食选择,推动外卖服务行业的可持续发展具有重要意义。Sugar-sweetened beverages have caused worldwide health concerns due to their contribution to high calorie intake and chronicle diseases.Against the backdrop of the rapid development of online food delivery platforms,this paper utilizes sales data from delivery services in six Chinese cities and employs machine learning methods to identify consumption patterns of sugar-sweetened beverages.Additionally,the paper analyzes sales variations in simulated scenarios such as price discounts and sugar taxes.The findings of the paper are threefold:First,neural networks within supervised machine learning models demonstrate greater applicability for predicting consumption trends compared to traditional econometric regression models.Second,model predictions indicate that a 10% price reduction leads to a 0.49% increase in total sales,suggesting significant price elasticity of demand for sugar-sweetened beverages.Conversely,a 10% price increase results in a 0.17% decrease in total sales volume,implying that the utilization of taxation as a strategy to reduce the consumption of these beverages is viable.Third,there are spatial and temporal variations in the sales of sugar-sweetened beverages on delivery platforms,with higher sales volumes in the Yangtze River Delta region compared to other areas,and weekday sales surpassing those of weekends,with afternoons being the peak period.This research is significant for establishing pathways and mechanisms to reduce excessive consumption of sugar-sweetened beverages,guiding consumers towards healthier dietary choices,and promoting sustainable development in the food delivery service industry.

关 键 词：含糖饮料 外卖 机器学习 神经网络 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] F724.6[自动化与计算机技术—控制科学与工程] F426.82[经济管理—产业经济]