基于电商大数据的农产品短期经营风险预测研究——以家庭经营梨果种植户为样本  

Short-term Agricultural Operating Risk Prediction with Big Data of E-commerce: A Case Study of Family-run Pear Farmers

作　　者：程欣炜 岳中刚 CHENG Xinwei;YUE Zhonggang(School of Economics,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京邮电大学经济学院,江苏南京210023

出　　处：《商业经济与管理》2022年第9期16-29,共14页Journal of Business Economics

基　　金：教育部人文社会科学基金一般项目“基于深度学习的农产品电商风险预警和金融对策研究”(18YJC790016);江苏省社会科学应用研究精品工程课题“脱贫攻坚阶段农产品电商对小农户有机衔接现代农业发展效率的影响研究”(20SYC-057)。

摘　　要：农产品电商具有农业经营改善和农户福利提升的积极作用,在电商参与模式下识别和预测农产品短期经营风险有助于农户家庭主动管理和积极干预电商嵌入结构,也可为区域电商布局优化提供理论依据和预警方案。文章以江苏省3755户家庭梨果种植户为样本,采集种植投入、电商交易和借贷保险等电子台账大数据,对比分析线性计量模型、一维非线性机器学习模型和二维深度学习模型对农产品短期经营风险结果和风险决策的预测能力。结果表明,电子台账大数据包含预测农产品短期经营风险的信息含量,基于特征灰度的二维卷积神经网络具有精准而稳定的预测能力,电商交易的预测信息含量高于种植投入和借贷保险且三者相互印证。在此基础上,从电商数据规范、台账软件升级、电商咨询建设和数字平台管理角度提出建议。Agricultural e-commerce has a positive effect on operation improvement and welfare enhancement. Identifying and predicting its risks helps to actively manage and intervene embedding patterns,and also provides theoretical basis and early warning solutions for regional e-commerce optimization. Electronic ledger data of production,transaction and finance of 3755 pear farmers in Jiangsu Province were collected to compare the predictive capability of linear model,1-dimensional non-linear machine learning models and 2-dimensional deep learning model on the risk outcomes and decisions of agricultural operations. The results show that ledger data have adequate information content to predict short-term agricultural risk. Convolution Neural Networks on feature grey scale has the best prediction ability. And the prediction information of transaction data is higher than production and financial data with mutually validated structure. Recommendations are made from the perspectives of e-commercial data specification,ledger software upgrading,consulting service construction and digital platform management.

关 键 词：农产品经营 风险预测 电商大数据 卷积神经网络 计算机视觉 

分 类 号：F713[经济管理—产业经济]