Algorithm Research Of Tag-Based User Interest Discovery In Weibo And Application

As representative network application formed from Web2.0era, MicroBlog has obtained a rapid development in recent years. Users of MicroBlog to post and receive micro blogs, which get fast forward diffusion through a large number of retransmission. A large number of users bring a large network and then a huge commercial value. So accurate discovering of the users’interest, which is thought of being able to enhance the experience of using MicroBlog, is very important. Due to the huge information, short length and noise data, the traditional method has poor effect in user interests discovering. We proposed a new method which use tag, bipartite graph algorithm and user interaction relationship diagram to improve the performance of current methods.First, after the analysis of the characteristics of the micro blog and defects of existing algorithms of discovering users interest, we use tags to represent users’ interest and put forward the user interest discovering algorithm based on bipartite graph, then use it to recommend tags to the users of MicroBlog. In order to solve the matrix sparsity problem in tag recommendation, we use tokenizer to split micro blogs to words and to build User-Tag and Tag-Word bipartite graphs respectively, and then we use the tags to descript users’ interest. Meanwhile we use users’interaction relationship to establish tag relationships, and propose a method called TagRank which based on PageRank, which uses tags to describe users’interest. Finally, we combine the two methods to find users’interest. We evaluate our method by a series of experiments based on a data set crawled from a famous Chinese MicroBlog website and analysis the results. The results show that our method performs better than traditional user interest discovering methods. After that, we apply our methods to build a users’interest discovering module in a practical systems-the trusted business social network system-to achieve personalized recommendation based on the users’interest.