A Novel Friend Recommendation Algorithm Based on Intimacy and LDA Model
Zhenwu Wang, Xiaohui Han, Tian Haoming · 2017
With the development of Internet, various social network service (SNS) platforms appeared, such as Facebook, twitter, Flickr, Sina microblog, and so on. Friend recommendation is the key issue for the SNS which can enhance the interactivity among SNS users.A novel recommendation algorithm is proposed in this paper, it applies time line to compute the interactions among target user and his/her recommended friends firstly, which predicts the intimacy trend and fits intimacy with interactive information at different time slots; then a Latent Dirichlet Allocation (LDA) model is used to generate subjects and judge the subject similarities between target user and recommended friends, at last, the two parts have been combined by an information entropy method which adjust the weight information dynastically during the friend recommendation process. Compared with collaborative filtering recommendation algorithm and LDA method, the experimental results proved that the proposed algorithm has got better performance.