摘要
传统的协同过滤推荐算法大部分只考虑单一的用户相似度,而忽略了用户其他特征,随着Web2.0和社交网络等互联网新概念模式的发展,用户对个性化推荐技术的要求越来越高.针对上述情况,提出一种结合社交与标签信息的协同过滤推荐算法.首先,定义了小众重叠度和个体重要度的概念,并描述了"个体-小众-社区"的形成过程;然后,分析"用户-项目-标签"三元组信息获得用户间的相似度,并结合社区中的个体重要度,最终得到目标用户的偏好预测和个性化推荐.采用Last.fm公共数据集进行一系列对比实验,实验结果表明,新算法在一定程度上提高了推荐准确度.
Most of collaborative filtering recommendation algorithms explore techniques for matching people with similar interests and making recommendations on this basis, while ignore the users＇ other features such as the social information. With the arising of the new concepts such as Web2.0 and social network services, the personalized recommendations are paid more and more attention. To combat the question, a collaborative filtering recommendation algorithm combining users＇ social information and items＇ tags informa- tion is proposed in this paper. Firstly, two concepts, the degree of overlapping between groups and the importance degree of individu- ality, are defined, and the constructing process of the ＂individuality-group-community＂ is illustrated. Then, the similarity between users is obtained through analyzing ＂users-items-tags＂ triples. Finally, combining the importance degree of individuality in the com- munity, the target users＇ preference predictions and personalized recommendations are made. A series of comparative experiments are done on the public data set of ＂last. fm＂. The results show that the algorithm is much better in the recommendation accuracy.
出处
《小型微型计算机系统》
CSCD
北大核心
2013年第11期2467-2471,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61073146)资助
重庆市自然科学基金项目(cstc-2009BB2082&2011jjA40045)资助
关键词
协同过滤
推荐算法
社交网络
标签
collaborative filtering
recommendation algorithm
social network
tag
