基于主题的微博二级好友推荐模型研究

随着社交网站用户爆炸性增长，寻找与自己兴趣相投的潜在朋友越来越困难。为了有效解决以上问题，基于社会关系理论中的同质性理论和三元闭包关系理论，分别从社会关系和内容两个维度向社交网络用户推荐志同道合的朋友。并利用LDA的扩展模型UserLDA对新浪微博用户进行兴趣主题建模，通过用户-主题概率分布矩阵计算用户相似度，以进行TopN二级好友推荐。在真实微博语料库上进行试验表明该推荐算法有较好的准确性和多样性。

Two-level MicroBlog Friend Recommendation Based on Topic Model

With users' explosive growth in the social network, it is more and more difficult for them to find potential friends of similar interests. In order to effectively solve the above problem, this paper proposes to recommend like-minded friends for social network users respectively from two dimensions of social relations and contents based on the theories of Homogeneity and Triadic closure. The paper models interested topics for Sina MicroBlog users by using extended LDA model-UserLDA, and calculates the users' similarity through user-topic probability distribution matrix, to recommend TopN two-level friend. Through the experiments on real weibo corpus, the result shows that the recommendation algorithm has better accuracy and diversity.