欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
协同过滤是推荐系统中广泛使用的推荐技术，对推荐结果可解释强。基于用户的协同过滤是一种重要的系统推荐方法，用户评分数据的极端稀疏性制约着系统的推荐质量。针对上述情况，提出一种基于用户群体影响的协同过滤推荐算法。首先，定义了用户群体的概念并根据群体影响提出两条相应准则；然后，计算用户相似性时，不仅考虑了用户个体之间的相似性，而且考虑了用户所处群体之间的相似性。该算法不仅可以更加精确地刻画用户之间相似度，而且一定程度上增强了推荐系统的稳定性。实验结果表明，该算法能有效地提高系统的推荐质量，而且满足所提出的两条准则。
英文摘要：
Collaborative filtering is a popular recommendation technology in recommender systems due to its strong interpretability to recommender results. User-based collaborative filtering is one of the important system recommender methods, but the recommended quality of systems is restricted by the extreme sparsity of user rating data. To solve this problem, this paper proposes a new collaborative filtering recommended method according to the influence of the user- group. Firstly, this paper gives a definition on the new term user-group, and then proposes two criterions according to the influence of user-group. Secondly, this paper calculates the similarity of users via considering not only the similarity among users but also the similarity among groups. The proposed method can precisely characterize the similarity among users as well as enhance the stability of recommender system to some extent. Experimental results on real datasets show that the proposed method efficiently improves the recommended quality of systems and satisfies the proposed two criterions.
同期刊论文项目
同项目期刊论文
期刊信息
位置：