【作者】 胡必云； 李舟军； 王君；

【Author】 Hu Biyun,Li Zhoujun,and Wang Jun (School of Computer Science and Engineering,Beihang University,Beijing 100191)

【机构】 北京航空航天大学计算机科学与工程学院；

【摘要】 基于邻居(neighborhood-based)的协同过滤是一项很受欢迎的用于推荐系统的技术.它可以分为基于用户(user-based)和基于项目(item-based)协同过滤.它通过用户或项目之间的相似性预测用户对于未评分项目的偏好.然而,传统的相似度方法易受数据稀疏影响.为了解决这个问题,提出了基于心理测量学(psychometrics-based)的相似度方法.实验结果表明,提出的相似方法更适合基于邻居协同过滤,它们可以提高推荐准确性和覆盖度(coverage).更多还原

【Abstract】 Neighborhood-based collaborative filtering is a popular technique used in recommendation systems.It can be further divided into user-based and item-based.It predicts a user’s preference for an unrated items based on the similarity between two users or items.However,traditional similarity methods are susceptible to data sparsity.To overcome the drawback,this paper proposes novel psychometrics-based similarity methods which are less susceptible to data sparsity.Experiments on public dataset show that the proposed similarity methods are better for neighborhood-based collaborative filtering,and they can improve the recommendation accuracy and coverage.更多还原