基于机器学习的食品安全风险预测与监管政策启示——以生鲜水产品为例  

Food Safety Risk Prediction and Regulatory Policy Implications Based on Machine Learning:Evidence from Fresh Aquatic Products

作　　者：周洁红[1] 魏珂 金宇[2] 徐子龙[2] ZHOU Jiehong;WEI Ke;JIN Yu;XU Zilong

机构地区：[1]浙江大学中国特色社会主义研究中心,杭州310058 [2]浙江大学中国农村发展研究院,杭州310058

出　　处：《农业经济问题》2024年第5期4-19,共16页Issues in Agricultural Economy

基　　金：国家社科基金重大项目“推进居民绿色消费升级的监管体系研究”(编号:19ZDA106);国家社科基金项目(编号:22VRC187);中央高校基本科研业务费专项资金资助;浙江省社科规划办重大项目“全面形成绿色生产生活方式研究”。

摘　　要：食品安全风险的准确识别与预测是提升现代化食品安全保障能力的关键。在当前食品安全风险成因复杂、监管靶向性不足的背景下,构建可靠的风险预警机制有助于我国监管部门摆脱监管资源投入较大、监管效率较低的困境。本文基于全国2014—2022年超30万生鲜水产品抽检样本,构建生鲜水产品风险的机器学习预测模型,并依据风险成因的重要性对各类预测特征进行排序,最后结合计量方法分析重要特征对生鲜水产品风险的影响。研究表明,随机森林模型对生鲜水产品风险的预测灵敏度与精确度达75.4%和78.0%,生鲜水产品风险与抽检发生的供应链环节、抽检地区、政府监管强度、水产品类别以及天气状况五个维度的特征紧密相关。这为实现我国食品安全风险的准确预测与成因分析提供了新的视角,也为优化政府食品安全监管资源配置提供了可靠依据。Accurate identification and prediction of food safety risks are crucial for enhancing modern food safety assurance capabilities.In light of the complex nature of food safety risks and the inefficiencies in regulatory processes,establishing a dependable early warning mechanism can help the Chinese government overcome the challenge of excessive investment in regulatory resources with low efficiency.This paper utilizes nationwide sampling data encompassing over 300,000 fresh aquatic products from 2014 to 2022 to construct a machine learning prediction model for assessing risks associated with fresh aquatic products.Various prediction features are sorted according to the importance of risk causes,and the impact of important features on the risks of fresh aquatic products is analyzed using econometric methods.The findings demonstrate that the random forest model achieves a sensitivity and accuracy rate of 75.4%and 78.0%respectively,in predicting risks linked to fresh aquatic products.These risks are closely intertwined with five key dimensions:supply chain links,regions,government supervision intensity,aquatic product categories,and weather conditions.This paper provides a new perspective for accurate prediction and cause analysis of food safety risks in China,and also provides reliable evidence for optimizing the allocation of government food safety supervision resources.

关 键 词：食品安全 风险预测 机器学习 监管效率 水产品 

分 类 号：F322[经济管理—产业经济] TP181[自动化与计算机技术—控制理论与控制工程]