作　　者： ; ; ; ; ;

机构地区： 中山大学信息科学与技术学院

出　　处： 《计算机学报》 2010年第8期1369-1377,共9页

摘　　要： 文中围绕传统的协同过滤推荐算法存在的局限性展开研究,提出一种不确定近邻的协同过滤推荐算法UNCF.根据推荐系统应用的实际情况,对于推荐的每一种场景其实都是不可预先确定的,而文中算法基于用户以及产品的相似性计算,自适应地选择预测目标的近邻对象作为推荐群,同时计算推荐群中推荐把握概率较高的信任子群,最后通过不确定近邻的动态度量方法,来对预测结果进行平衡的推荐.通过实验结果表明,该算法可以有效平衡用户群以及产品群推荐结果所带来的不稳定影响,有效缓解用户评分数据稀疏的情况所带来的问题,并在多个实验数据中,提高了推荐系统的预测准确率. To overcome several limitations in the research area of collaborative filtering（CF）,this paper presents a CF recommendation algorithm,named UNCF（Uncertain Neighbors＇ Collaborative Filtering Recommendation Algorithm）.In the reality,the scene of recommendation is uncertain.The similarities computations of both user-based and item-based are considered to choose the neighbors dynamically as the recommendation set.This set can be used to select the trustworthy subset which is the most effective objects to the predicted result.Moreover,this paper defines a new prediction algorithm that combines the advantages of trustworthy subset for this uncertain recommendation method.Through experimental results,the UNCF algorithm can consistently achieve better prediction accuracy than traditional CF algorithms,and effectively leverage the result in the uncertain environment.Furthermore,the algorithm can alleviate the dataset sparsity problem.

关 键 词： 不确定近邻 协同过滤 推荐系统 相似性度量 信任子群

领　　域： [自动化与计算机技术] [自动化与计算机技术]