Combining geographical information of users and content of items for accurate rating prediction
Zhi Qiao, Peng Zhang, Jing He, Yanan Cao, Chuan De Zhou, Li Guo · 2014
Recommender systems have attracted attentions lately due to their wide and successful applications in online advertising. In this paper, we propose a bayesian generative model to describe the generative process of rating, which combines geographical information of users and content of items. The generative model consists of two interacting LDA models, where one LDA model for location-based user groups (user dimension) and the other for the topics of content of items(item dimension). A Gibbs sampling algorithm is proposed for parameter estimation. Experiments have shown our proposed method outperforms baseline methods.