Research And Application Of Food Safety Risk Prediction Method Based On Sample Expansion

Food safety is related to people’s health and social economy,and its importance cannot be ignored.Food safety incidents,led by the recent spread of the novel coronavirus through frozen food,have caused serious losses to the global economy,and it is urgent to strengthen the ability to predict food safety risks.At the same time,due to the small effective sample size of some food safety risk data,the prediction accuracy of the data-driven risk prediction model is insufficient,and effective prevention and control measures cannot be provided.In this context,how to improve the accuracy of risk prediction to avoid food safety risks is of great significance to the sustainable development of society.This paper takes sterilized milk as the object to carry out research,and explores the impact of risk indicators on the whole,the effective data expansion methods and the risk prediction methods.It can provide certain technical sustain for relevant departments to formulate food safety risk prevention and control,and has certain significance for the implementation of food safety social co-governance.The main elements of this article are as follows:1.Aiming at the problem that the traditional single risk assessment method assigns the weights of food safety factors one-sided,which leads to the inability to comprehensively assess the food safety risk,a food safety risk assessment method based on the coefficient of variation-entropy weight(CV-EWM)is proposed.The entropy weight method and the coefficient of variation method are used to assign weights to the food safety risk impact indicators respectively.According to the principle of minimum information entropy,the indicator weight value is fused and the comprehensive risk value is calculated,so as to divide the overall risk level of food safety according to the risk value according to the five-scale method.2.Aiming at the problem that the effective sample data of food safety risk is small and the accuracy of predicting food quality risk is not high,a sample expansion method based on Monte Carlo(MC)and ADASYN(MC-ADASYN)is proposed.The MC is used to expand the effective data of food safety,and the U test is used to verify the credibility of the expanded data.Then combined with the CV-EWM weight distribution method to obtain the risk level of the expanded food safety risk effective data,and then use the ADASYN method to solve the imbalance of the expanded data,thereby providing sufficient and reasonable training data for the improvement of the accuracy of the food safety risk early warning model.3.Aiming at the problem that the internal parameters of random forest algorithm(RF)can’t be adaptively selected,which affects the accuracy and convergence speed of the constructed risk prediction model.And combined with the extended food safety and effective data,an improved random forest(RF)risk prediction method(CS-RF)based on cuckoo algorithm(CS)is proposed.Based on the effective sample data of food safety extended by the MC-ADASYN method,the optimal internal parameters of RF are found by CS adaptively,and an early warning model of food safety risk based on improved RF is constructed.The proposed method was applied to the food safety testing data of sterilized milk,and the validity and practicability of the constructed model are proved by performance evaluation indicators.And analyze the effective data early warning results of sterilized milk food safety,and provide food safety risk prevention and control measures for food testing institutions and enterprises.