Research Of Mining On The Public Opinion Information Based On Social Network

With the rapid development of computer network science and social network,the number of network users is growing exponentially,so the relevant data such as users on the network,events,public opinion mass increases.Some network media,such as sina,sohu,BBS and microblog on other large web,become the main platform to spread and gain information.As the fast development of information construction,government policy makers and the managers of relevant functional departments demands for direction of public opinion of the network media,so the discovery and management of network opinion become a problem to be solved.Therefore,the related research and practice on the online public opinion management has made great progress in recent years,and the present study is mainly about the research of network public opinion information that has its various data source,no unified data size.Foreign scholars mainly focus on social media,research methods such as tweet.This research focuses on China's weibo social media public opinion information organization and management and data mining,which mainly follows life cycle structure of network public opinion and the law of the development process,based on theory method of the network audience groups view.The research content of this article is divided into the following contents:(1)Information search and extraction research based on social network public opinion.We clearly define the concept of social network public opinion information,and analyze the scope,source,characteristics of network public opinion information.On this basis,the data sources is from namely social network media to sina weibo's public opinion multiple columns and the access way is to crawl or API port.At the same time,we search and organize public opinion information according to the symmetric word way theme of the network public opinion events.The article establish public opinion retrieval methods of personalized search behavior audience and public opinion information searching,cleaning,extraction,storage method,which provide an effective method to the social public opinion information organization and management.(2)Modeling and mining of social network public opinion group audience preference.Network subject interests and value orientation often become the important premise of network public opinion.Through the analysis of the microblog content and the Weibo user preference model from individual point of view in the events of public opinion,it aims to control and grasp attention degree of the Weibo user personality characteristics.This paper establishes the representation of text feature vector about micro-blog information;establishes micro-blog text feature vector space by micro-blog and the corresponding weights of the feature words;establishes a multi classification text SVM model of public opinion groups audience preference.(3)The discovery of opinion leaders in social network public opinion events.Opinion leaders is in an extremely important critical stage in the life cycle of social network public opinion.Relationship network is established between the audience comments text similarity and emotional similarity.Accrording to the cumulative points for each node in the framework of the system,and the integral value of the highest audience is the most influential opinion leaders' speech.According to the latest column topic structure of the social network,we construct network opinion leader mining model of the independent theme and complex cross theme.The validity of the method of mining process in the social network public opinion leaders is verified through experiments.(4)Research on community opinion and community structure in social network public opinion.In this paper,LDA algorithm is used to mine the sentiment and the regional distribution of the comment text,then the community opinion is generated and the cross entropy method is adopted to community structure mining.In the establishment of community opinion discovery model,the emotional statement tree method is proposed to mining the largest text information statement,which is used to refine subject audience opinion words through topic model iteration.Finally,the effectiveness of the social network group opinion and structure mining model is verified by experiments.