The Research On The Recommendation Algorithms For The User’s Tags And The Followees In Micro-blog

In the Internet era, micro-blog, as a new social media, is changing the way people communicate and get information. However, with the explosive growth of micro-blog users, micro-blog data also increased radically. People are facing the growing serious problem of information overload. Recommendation system is one of the effective methods to help users to discover the useful information and overcome the problem of information overload. How to dig out the useful information for the users from the massive data is an important application of recommendation system in the micro-blog platform.Based on the characteristics of the currently popular micro-blog platform, two recommendation algorithms for the micro-blog users are proposed in this paper. The main contents of our work can be summarized as follows:(1) In view of the fact that there have not enough tags for the most users in Sina micro-blog website, an algorithm based on RBLDA model and users’interaction graph for tags recommendation is proposed in this paper. Combining the topic model with social networking topology, the algorithm takes two kinds of relationships into consideration, that are friend relationship and interaction relationship. Firstly, the algorithm recommends the initial tags, which utilizes the RBLDA model to dig the topics of tags from the users’ friends. And then we consider the interaction graph, tags are transferred at a certain probability in the interaction graph. Finally, we rank all of the tags and choose the top k tags as users’ final recommendation tags.(2) For the issue that how to recommend the followed friends whom the micro-blog users may be interested in, an algorithm based on matrix factorization model is proposed in this paper. The micro-blog data present diversity. Each data source has reflected the user’s interests and preferences in different degrees. We extracted the characteristics of user’s interest from each data source, and then introduced into the matrix factorization model which is suitable for recommending followed friends. Finally, we optimize the model and get the best factor parameter matrix to recommend followed friends.The experiments carried on the real Sina micro-blog data sets show that the two algorithms are both feasible and effective. Compared with the traditional recommendation algorithms, the two proposed algorithms improve a lot in the aspect of accuracy.