Research On Early Warning Model Of Meat Product Quality Safety Based On Bayesian Network

PURPOSE Food safety issues have always been the focus of attention and research in China and the world.Sensitive and hot topics such as food safety will cause panic and mistrust of the government,endanger people ’s physical and mental health,and affect the credibility of the country.Therefore,it is very necessary to establish relevant early warning models.This article focuses on meat products and establishes a risk warning model for the safety of meat products,so as to have a more comprehensive monitoring and prediction of the risk status of meat products.The method and experience of early warning analysis of the safety risk of a certain type of food can be extended to all food categories.METHOD This article briefly explains the current status of food and meat product safety issues at home and abroad,introduces related concepts and methods,and provides a more complete introduction to the focus of this study,the Bayesian network.In the case analysis stage,an early warning model of meat product safety risk was established.First,use the 2015-2017 meat product sampling data provided by Food Safety Social Co-governance Project(SQ2017YFC160082),after preprocessing,based on the improved entropy method,using the risk index defined in this article,finally the complete modeling data of the early warning model is calculated.After that,the Bayesian network risk early warning model of meat product safety risk was established using Ge NIe2.0 software.Use testing indicators and full chain indicators,such as types of meat products,sampling location,unit level,sample type,packaging classification,unit type,sampling conclusion,area type,whether it is a capital city or capital,economic zone,production date,inspection date,sampling quantity,shelf life,etc.for Bayesian network structure learning and parameter learning,get the final risk early warning network model.RESULT The conditional probability distributions of meat product safety risk levels from level 1 to level 5 are safety(41%),light police(30%),intermediate police(16%),serious police(6%)and special police(7%).The network has 31 nodes,121 states and 9070594 parameters.After verification by the verification set,the accuracy of the model is 73.01%.After verification by the regional data,the model prediction results are in line with the actual situation.After sensitivity analysis,there are 6 variables in the network as key risk factors.The model has high visualization degree,strong self-learning ability,excellent performance in predictive analysis and reasoning,and has good practical application significance.CONCLUSION This model can effectively realize the early detection,early understanding,early decision-making and early intervention of the safety risk status of meat products,referred to as "four early".Different policies and measures are adopted according to the different risk levels,so as to achieve the purpose of early intervention and avoid the occurrence of meat product safety incidents.This model can improve the current status of food supervision workers and point out a direction for future research on food safety risk early warning.It is necessary to enter the demand-oriented fullchain social co-governance application early warning research and make meat product safety and food safety the norm.