Abstract: The data sparseness is due to the nature of traditional collaborative filtering and trust-based recommender systems can effectively deal with the sparse data without losing accuracy. It is appropriate to use different methods for different users to give more personalized recommendation. The vertex characteristic in microcosmic stratums was studied, and the formal definition of interest was proposed. It was used to demonstrate the impact of local structures of the recommended user on the results of recommender systems. In the end, several results were given to illustrate the diversity of the effects of recommender systems on users of different types.

Key words: trust, recommender system, local network structure, interest, personalized recommendation

摘要： 结合信任的推荐系统可以有效地缓解传统协同过滤算法中存在的数据稀疏问题,并能给每个用户提供可信且准确的推荐。然而系统中的每个用户都是不同的,因此考虑针对不同用户应采用不同推荐模式来查找推荐群体,以做出更具个性化的推荐。研究了微观层次上的节点特性,引入了兴趣的概念,证明了被推荐者的多种节点特性对于推荐结果的影响效果。最后通过多组实验验证了推荐系统在具有不同特性的节点上的推荐效果差异。

关键词: 信任, 推荐系统, 局部网络结构, 兴趣, 个性化推荐