Research On Tag-Topic Identification And Community Mining In Social Network

The rapid development of Web2.0 technology and social media provide people an open and convenient platform to get information,express opinions and interact with others.The massive and complex information is generated in this social network,which is hidden behind potential friends,opinion leaders,hot events and other useful information,and the key to obtaining these useful information is to find the community of users with similar characteristics.The implementation of effective community mining not only affects people's production and life,but also plays a very important role in promoting the harmonious development of society.The information of social network mainly comes from the text content of user generation and interaction.Therefore,community mining is no longer a discovery of a single network structure,but more need to focus on the semantic understanding of these text information and the mining of content.Based on the information organization model of social tag system,the user relationship is used as the research object.Based on identifying the topic of the user tag and the subject of the network content,the structure,subject and knowledge of the community are excavated.1)Propose a method of user importance discrimination based on tag topic.The interest of the user may be multifaceted and the traditional method does not distinguish the "multi interest"of the user.It causes the user's importance not to be carried out in the same interest category of the user which result in the problem of "interest deviation" in the calculation of user similarity.To solve this problem,this paper firstly uses tag clustering method to identify interest topics and classify users in Folksonomy mode network.And then,it constructs user importance index with social network analysis and PageRank method in the user community with the same interest topic.Finally,it introduces the index into the user similarity model,verify the validity of the delicious dataset and apply it in friend recommendation.2)Propose a user interest renewal model based on tag topic.The establishment of a user interest model for social networks is a great significance for providing high quality network personalized services,and the identification of changes in user interest which is a difficult point in modeling.In the view of the poor characteristics of the social networks which have not been built in the Folksonomy mode,the "words" in the LDA topic model are used as tags,and the semantic features and time characteristics of the tags are integrated to build the user interest renewal model.First,according to the difference of micro-blog information,users are divided into two categories,and the suitable method is used to construct the model respectively.For the old users with rich micro-blog information,the time weighting function is introduced to build the user's LDA interest renewal model,and the similarity degree of space vector is adopted for the "cold start" users with less micro-blog information.The user interest model is obtained by volume method,and the user interest model is updated by learning model to identify the change of user interest.The proposed method is applied on the micro-blog data set,and the topic of the network,the core user of the network topic,and the interest of the user are obtained.3)Put forward a comprehensive consideration of user social relations and user generated community partition method.This section consists of two stages:the determination of user similarity and community partition based on information granularity.From two perspectives of user social relations and user generated content,the first phase optimizes the social relationship model by using the link prediction method.Using 'fine grained' user tags and 'coarse-grained'content tags to build a user tagging-topic relationship model.The two models are weighted together and set adaptable.An integrated user similarity model is established based on the adjustment parameters of social relations and user content.In the second stage,in view of the shortage of K-Means clustering algorithm and the high dimension and sparsity of the data,the information granularity principle is applied to the user clustering analysis and the membership degree and the generalized equivalence relation of the user equivalence relation are given.On this basis,a social community partition algorithm based on the information granularity is proposed.The experimental results show that,by comparing with the unweighted user tag topic model and K-Means,the proposed model has obtained better evaluation result of I index and Dunn index than the unweighted user tag topic model and the information granularity method.4)Propose a community's knowledge growth measurement and user selection method.The social tag system is applied to the research of knowledge service.Using the key technology and research results of the previous community mining,a semantic knowledge base in the model of mixed tag ontology is established to analyze the composition and characteristics of the generated knowledge community.Quantified express the knowledge transfer between individuals and organizations and measure knowledge stock and knowledge growth in knowledge community.From the point of view of content perception,we combine the similarity calculation model to design the user selection algorithm of knowledge transfer process,optimize knowledge transfer and then promote the good and efficient flow of knowledge in the community.