Abstract: The trust based recommender system is to use the trusted entities to recommend items. As trust is a complex concept, to propagate and predict trust is an important task. A Statistical Relational Learning（SRL） model, Markov Logic Networks（MLNs）, is proposed to present the transfer properties of trust. The theory model of MLNs is discussed. With MLNs’s reasoning algorithm, the trust relationships are predicated. In the trust based recommender systems, the experimental results show that MLNs has a higher accuracy and better solution of cold-user problem than MoleTrust approach.

Key words: Markov logic networks, trust, recommender systems, statistical relational learning

摘要： 基于信任的推荐系统是利用信任的实体进行项目推荐，然而信任是一个复杂的概念，对信任进行传播和预测是一项重要的任务。提出了用一种统计关系模型——Markov逻辑网来表示信任的传递性质，讨论了Markov逻辑网的理论模型，通过其推理算法预测信任关系，实验结果表明，在基于信任的推荐系统中Markov逻辑网方法比MoleTrust方法在推荐精度和解决冷用户问题上有更好的效果。

关键词: Markov逻辑网, 信任, 推荐系统, 统计关系学习