Abstract: Collaborative Topic Regression（CTR） combines ideas of probabilistic matrix factorization and topic modeling for recommender systems, which has obtained successes in many application. But this model does not take into account the impact of user social relationships on user interests. In order to solve this problem, the probability link is incorporated in this algorithm to excavate the influence of social relation network on the latent structure for users, and it`s also in constraint on the object function. This paper proposes a novel hierarchical Bayesian model called User Social Relation incorporated into Collaborative Topic Regression（USRCTR）, which extends CTR by integrating the user-item feedback information, item content information, and social relation network. Experiments on the Lastfm dataset show that this model can achieve better prediction accuracy than several improved methods of CTR with lower empirical training time.

Key words: recommender systems, collaborative filtering, topic models, social relations network

摘要： 协同主题回归（CTR）将概率矩阵分解和主题模型结合应用于推荐系统，在许多推荐应用中取得了成功，但该模型没有考虑用户社会关系对用户兴趣的影响。针对该问题，引入概率链接函数来评估社会关系网络对用户兴趣的影响，并以此约束目标函数。在CTR的基础之上，提出一种融入用户社会关系的协同主题回归模型（USRCTR），结合用户项目评分信息、项目内容和社会关系网络，构建一个基于分层贝叶斯模型的推荐引擎。在Lastfm数据集上实验表明，与其他几种CTR改进方法对比，该模型的训练时间更短，推荐精度更高。

关键词: 推荐系统, 协同过滤, 主题模型, 社会关系网络