摘要
针对目前个性化服务中用户模型稳定性低、推荐结果不尽人意的现状,在建立基于本体的用户兴趣模型基础上,通过模型更新提高稳定性,建立用户群实现用户模型管理。提出利用矩阵聚类降维分解技术的个性化推荐算法,引入偏好方差的概念计算用户最近邻,进而产生推荐,避免了传统协同过滤算法的数据稀疏性缺陷,提高了推荐质量。结合面向电影的个性化推荐系统,验证了模型及算法的有效性。
To deal with low stability of user model and unsatisfied recommendation result existing in personalized service nowadays,ontology-based User Interest Model（UIM） was set up.Model stability was improved by update and model management was realized by establishing user group.To avoid sparse data in traditional collaborative filtering algorithm,personalized recommendation algorithm utilizing matrix clustering dimensionality-reduction decomposition was proposed.Nearest neighbors were calculated according to preferences variance,and the recommendation could be obtained subsequently.Then,the recommendation quality was improved.Finally,effectiveness of the model and algorithm was proved through a personalized movie recommendation system.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2010年第12期2757-2762,共6页
Computer Integrated Manufacturing Systems
基金
国家自然科学基金资助项目(61073090)
上海市科学技术委员会科研计划资助项目(09DZ1122302)
广东省教育部产学研结合资助项目(2009GJE00026
2009B090300429)
上海市重点学科建设资助项目(B004)~~
关键词
本体
用户兴趣模型
稳定性
数据稀疏
偏好方差
ontology
user interest model
stability
data sparseness
preferences variance
