Fusion Of User Tags And Microblogging Content For Users Interested In Community Discovery

With the continuous development of social networks, microblog has become an important and indispensable part of people’s daily lives. In the platform of microblog, the user label defined by users and the activities of deploying and retweeting microblogs did by users can reflect the interests of user, and how to use these information, mining the interests of users and finding the user interests communities is an important job of great research significance and value. In this paper, we did deep research and discussion about the user interests communities finding of users of microblog at the following several aspects:(1) A user label interest model construction method based on feature mapping is proposed.According to the characteristics of user labels that can reflect the interests of users, we choose user labels as the features of user interests model, and in order to solve the problem of data sparseness and noise influence caused by different label expression and long labels, the feature mapping theory is introduced, the long labels is divided into sub-labels set by segmentation, and by calculating the similarity of labels, mapping the user label to the feature label of the max similarity, and use the product of label similarity and label frequency as the value of the feature dimension, to construct the model of user label interest, and we use the fuzzy clustering method to validation the effect of the user label interest model.(2) A user microblog content interest model construction method based on supervised LDA is proposed.According to the influence to the topic distribution of microblog caused by the interaction of microblog text, a microblog generation model based on supervised LDA is proposed, by combining considering the four aspects including microblog retweet, microblog comment, microblog reply and others’comments that affect the user microblog interest distribution, based on the traditional LDA model, we construct the supervised LDA microblog generation model, and then get the topic distribution of microblogs, and further more we get the topic distribution of users.(3) A user interest community discovery method combining user labels and microblog content is proposed.Based on the researches of (1) and (2), using the similarity of user models to construct user label interest relationship network and user microblog content interest relationship network,fusion them with the existing user relationship network, and based on the nework, considering the community overlapping problem caused by the fact that a user belongs to many communities, we proposed a user interest community discovery method based on k-clique,by solving the community overlap matrix,we get the community connect matrix.and finally we can get the user interest community which contains some connected k-cliques.(4) Based on the achievements of researches mentioned above, we designed and realized the prototype system of microblog user interest community discovery.