贝叶斯网络及其在白酒安全预警中的应用

Bayesian Network and Its Application in Liquor Safety Early Warning

摘要:

近年来，白酒安全问题日益凸现，影响人们身体健康和白酒行业发展，白酒安全预警日益重要。本文以食药监局的监测数据为样本，研究基于贝叶斯网络的白酒安全预警方法。从可能影响白酒质量安全的金属污染物、农药残留、食品添加剂、品质指标、微生物污染和非食用物质6个方面的因素进行分析，划分食品安全状况等级与预警指标，运用最大似然估计算法和贝叶斯网络建立白酒食品安全预警模型，使用MATLAB软件进行仿真实验，对白酒危害因子的风险值和白酒食品安全的风险程度分类预测。结果表明，贝叶斯网络在白酒食品安全风险预警中具有较高的准确率，是一种能准确、稳定实现白酒食品安全风险预警的算法，有助于为政府监管提供决策支持，丰富食品安全预警技术方法。

Abstract:

In recent years, liquor safety has become one of the main factors, affecting people's health and the development of the liquor industry, and the liquor safety early warning is increasingly important. In this paper, the monitoring data of the Food and Drug Administration was used as the texting sample to study the liquor safety early warning based on Bayesian network. Firstly, six factors including metal contaminants, pesticide residues, food additives, quality indicators, microbial contamination and non-edible substances that might affect the quality and safety of liquor were selected as the assessment indexes, and the food safety status and warning indicators were divided. Then, the liquor safety early warning model was established by maximum likelihood estimation algorithm and Bayesian Network. The MATLAB software was used to simulate the experiment for predicting the risk value of liquor hazard factors and the degree of liquor safety risk. The results showed that the Bayesian network had a high accuracy in liquor safety risk early warning of, and it was also an accurate and stable method to realize the risk early warning of liquor, which could provide decision support for government supervision and enrich the early warning technical methods for food safety.