Research On Network Public Opinion Early Warning Model And Its Application For Food Safety

Even today,with the rapid development of science and technology and social progress,food safety is still the key issue of people’s livelihood.Food safety incidents are infectious,which have different degrees of impact on people’s lives.The inflow of unqualified food into the market not only brings harm to people’s personal safety,but also produces public opinion on the Internet and its disordered circulation,which is more likely to lead to social panic.Based on the analysis of the research status of food safety and related network public opinion at home and abroad,it is found that the traditional network public opinion analysis model is inefficient,the food safety risk early warning model is complex and the prediction accuracy is not high,which causes the current food safety information and risk communication mechanism to run smoothly.Focus on these problems,this paper takes the food safety related network big data as the main research object,analyzes and studies the network text analysis model and the food safety network public opinion risk early warning model,and develops a unified food safety information platform combined with the multi-dimensional data related to food safety,so as to further strengthen the exchange of food safety information and risk in China.In the above research,the main achievements are as follows.(1)Aiming at the problem of unbalanced distribution of all kinds of data sets in network text big data,this paper proposes a text feature extraction model based on the vAriance of Document Frequencies(ADF)and four text feature weight calculation methods based on the proposed model.Through machine learning algorithms such as random forest,support vector machine and naive Bayes,it verifies the effectiveness of the proposed model in the classification of unbalanced network data sets.Experimental results show that the improved text feature model based on ADF can achieve better classification effect.(2)Aiming at the streaming data characteristics of network text,this paper proposes an online and offline hybrid computing framework for text analysis based on Storm.Compared with the traditional text analysis computing mode,this framework can achieve higher accuracy and efficiency for the analysis and processing of streaming text big data,which verifies the effectiveness of storm framework in the field of text analysis.(3)Combined with the domain characteristics of food safety network text,through the establishment of mapping relationship among text,food safety keywords and cluster,the food safety network event detection algorithm based on singlepass text clustering algorithm is constructed;in addition,aiming at the big data characteristics of network text data,the event mining algorithm parallelization scheme based on storm is proposed,which effectively improves the accuracy and efficiency of the event detection algorithm.(4)Based on the network event index system model of hierarchical structure,the risk early warning index system of food safety network event is constructed.A risk early warning model called AHP-LSTM using long-term and short-term memory network(LSTM)based on analytic hierarchy process(AHP)is proposed.The model takes the index system data as the input,and uses the optimized analytic hierarchy process(AHP)to fuse the index value as the expected output of the model Finally,the public opinion risk early warning of food safety network events is realized.Through the comparison,it is found that the accuracy of the AHP-LSTM model proposed in this paper is higher than that of the traditional early warning model,which can more effectively predict the development trend of food safety events,and verify the effectiveness and reliability of the model.(5)Based on big data storage(Hadoop)and distributed computing(storm)framework,the food safety big data computing framework is studied and established to solve the problem of real-time analysis efficiency of food safety information under the background of big data.At the same time,combined with the four-dimensional data of food safety network public opinion data and detection data,a perfect food safety information platform is constructed to enhance the information and risk exchange among the main bodies of food safety.In view of the low efficiency of big data retrieval brought by the traditional relational database in the information platform,a database with non relational database mongodb as the data integration query platform is proposed,With the help of its weak relationship and the query advantage of massive data,the page response speed of the information platform is greatly improved.