Food Safety Risk Warning Based On Comprehensive Analysis Of Multi-source Data

Food safety risk early warning is of great significance for timely detection of potential food safety hazards.Due to the randomness and suddenness of food safety risks,that is foods may have unexpected situations due to information such as time and place,therefore China has always advocated the four-in-one long-term supervision system of "industry supervision,business self-regulation,consumer participation,and social supervision".On this basis,this article mainly researches the method of risk early warning in food circulation,and the main works are as follows:Analyze the data source subjects that can assist in food risk early warning during food circulation,and propose a comprehensive analysis method based on Weibo public opinion data,news media report data,and food sampling test data.Combining early public opinion discovery and mid-term risk assessment to form a multi-source food data risk early-warning system in the food circulation process,which for each part:1.Research on sentiment analysis methods of Weibo public opinion dataBefore specifying a sampling plan,public opinion analysis is needed for foods with many negative emotions and relatively harmful effects.This article first researches Weibo advanced search crawlers,and can obtain Weibo data containing specific food keywords in a certain period of time and area.Then,the sentiment analysis method based on Bert model is used to analyze sentiment data related to fixed category foods crawled on Weibo,and compared with deep learning models based on word2 vec combined with LSTM or Bi-LSTM.2.Research on text clustering methods in news media reportsThe collection of news reports on food incidents can also serve as a warning to the regulatory authorities in a certain area.Therefore,this paper proposes the interval between texts based on the traditional vector space model and combine the food and place naming entities in the news report to calculate the similarity,and finally use Single-Pass algorithm to cluster the food news text.This paper first uses a web crawler to crawl some representative food safety events from previous years and conduct simulation experiments to verify the algorithm's effectiveness.It is found that the algorithm proposed in this paper has improved the clustering effect.Finally,it is proposed to use the number of texts in the news cluster within a unit time period to evaluate the danger degree of a food incident topic,and to analyze the food incident reports of Guizhou Province collected on the platform in 2018 to perform algorithmic analysis on the relevant foods in the region early warning.3.Research on risk assessment methods of food sampling dataPublic opinion emotional analysis and text clustering of news media reports can only obtain the food categories involved in food safety incidents with high negative emotions or media reports,and basically only provide assistance for the sampling programs and plans of the relevant supervision and sampling departments,which plays an early warning role.In the actual situation,the relevant departments need to carry out targeted sampling of the relevant food,but how to analyze and risk assess the sample test results is also an important part of the risk early warning in the circulation link,so this paper considers the possibility and severity of the sampling results,based on the dual risk level method,the risk assessment of the sampling results is proposed,so as to better analyze the risk level of the sampling results and play an early warning role.