Abstract: In the recommendation system, user cold start problem is the problems existing in the traditional collaborative filtering recommendation system. Aiming at this problem, on the basis of collaborative filtering algorithm, a new hybrid collaborative filtering algorithm is proposed to alleviate cold start, when calculating the user similarity is introduced into the algorithm users trust mechanism and demographic information, comprehensive considering attribute similarity and trust of users. At the same time, this algorithm improves the neighbor selection of users. Experiments show that the proposed algorithm can effectively alleviate the cold start problem of traditional collaborative filtering recommendation system.

Key words: collaborative filtering, recommendation system, cold start, trust mechanism, user cluster

摘要： 在推荐系统中，用户冷启动问题是传统协同过滤推荐系统中一直存在的问题。针对这个问题，在传统协同过滤算法的基础上，提出一种新的解决用户冷启动问题的混合协同过滤算法，该算法在计算用户相似性时引入用户信任机制和人口统计学信息，综合考虑用户的属性相似性和信任相似性。同时，算法还在用户近邻的选取上做了一些改进。实验表明该算法有效缓解了传统协同过滤推荐系统中的用户冷启动问题。

关键词: 协同过滤, 推荐系统, 冷启动, 信任机制, 用户聚类