Food Safety Early Warning Analysis Method Rearche And Application

As a Chinese saying goes, food is the paramount necessity of the people whereas safety is the prioritized concern of food. Food safety, which has close connection with everyone’s life and health, is regarded as the most important fundamental and the guarantee of people’s livelihood. Through the analysis and exploration on the existing food safety management model, food safety information, monitoring data early warning analysis method and public opinion early warning analysis method both at home and abroad, this paper points out that there is asymmetric information among the stakeholders in the field of food safety. The stakeholders include government, food enterprises (manufacturer, transporter and restaurant), testing agencies, media and consumers. The asymmetric information does not facilitate the effective early warning analysis and supervision. The traditional early warning methods and technologies only focus on single-sourced early warning analysis and the effects are far from the actual demand. Early warning method and technology is still a bottleneck that hinders the effective supervision of food safety. In this sense, on the basis of the multi-sourced data, this paper conducts an all-rounded analysis over the food safety issue in Guizhou province. The data collected mainly come from more than 60,000 quantitative testing food samples,240,000 rapid screening food samples and 3 million data items about food safety in Guizhou Province from 2008 to 2015. Through analyzing related data from food enterprises, governmental agencies and public opinions, this paper has examined the following research contents and made innovations listed below.1. Food safety multi-sourced data collection study methodWe conduct systematic analysis over the data collection method of the stakeholders, such as testing agencies, government, enterprises, the public and the media and we also realize the effective multi-sourced data collection and exchange. Given the numerous non-structural data from the testing agencies, we propose a novel idea of triple-based structural knowledge description method and subsequently set up a multi-tier model which realizes the transformation of food standard from non-structure to structure. We adopt this model to construct food safety standard structural knowledge database. After combining the standard controlling method in the testing procedure, structuring of testing data across regions and effective data collection can be realized in different testing agencies.2. Improved association rules and improved food safety early warning method combining entropy-weighting AHP-BP neural networkThrough the analysis over the features of food safety testing data, combining the related knowledge about food safety and in view of the problem of rationality of periodic large-scale sampling and food safety early warning, we propose a novel idea of improved association rules and improved food safety early warning method combining entropy-weighting AHP-BP neural network. This method adopts the improved association rules to reduce the associative test index of the food of the same category. The reduced test index is to serve as the input end of the BP neural network. Additionally, the improved entropy-weighting AHP method is adopted to integrate the reduced test index with risks. Three value risks are generated, namely heavy-metal pollution risk, chemical pollutant risk and pathogenic contamination risk, which can serve as the output value of the BP neural network. Finally, the BP neural network is trained by using input and output. Thus, the risk early warning model is obtained.3. The study of food safety early warning analysis methodFive major factors affecting food safety risk are examined, including AVE, STA, OUT, OUD and MAX. The food safety early warning risk index is expressed structurally and a sampling method is proposed.With regards to the public opinions among the internet users, an improved k-means clustering and mutual information analysis method is proposed, with the focus on the spotlights and affective analysis of the public opinions. This method "mainly revises the density-peak algorithm and it determines the dc value by using binary search algorithm. The hot words are extracted by using the mutual information and the score of positive words and negative words in each clustering is calculated. Thus, the holistic emotional trend can by analyzed so that it can facilitate the decision-makers to make the justified response to the public opinion.4. Application of Food Safety CloudBased on the data-driven food safety early warning analysis method, an innovative food safety big data cloud platform is designed, which integrates multi-sourced data collection, storage, management, analysis and application. In dealing with the incident of "distorted news about food containing Formaldehyde of Guizhou province" in June 2015, this cloud platform successfully realizes the information exchange among government, enterprises, testing agencies, consumers and media and provides technical support for the shared-governance of food safety.