针对传统协同过滤算法存在的两个弊端:一是传统的相似性度量方法在评分矩阵稀疏的情况下很难准确地反映用户间的相似性,二是不能及时反映用户的兴趣变化,提出一种新的相似性计算方法.此方法把基于用户兴趣度的相似性度量与基于项目相似度的数据权重结合,形成一种考虑用户兴趣变化的相似性度量方法.实验结果表明,改进后的算法集成了上述两种方法的优点,对传统算法中存在的两个弊端进行了改善,在推荐准确度上有所提高.%Existing collaborative filtering algorithms have two drawbacks:the one is that the conventional similarity measurement is difficult to accurately reflect the similarities among the users when the scoring matrix is sparse; the other is that it can not reflect the users' interests change in time.Aiming at these drawbacks,a novel similarity measurement algorithm is put forward.In this algorithm,a similarity measurement method considering the change of users' interests is formed by combining the users' interests-based similarity measurement and the item similarity-based data weight.Experimental results show that the improved algorithm integrates the advantages of the two methods above-mentioned,and ameliorates the two drawbacks in traditional algorithms,which leads to the improvement in recommendation accuracy.
展开▼
