融合标签特征和时间上下文的协同过滤推荐算法  

Collaborative Filtering Fusing Label Features and Time Context

作　　者：窦羚源 王新华[1,2] 孙克[1] 

机构地区：[1]山东师范大学信息科学与工程学院,济南250014 [2]山东省分布式计算机软件新技术重点实验室,济南250014

出　　处：《小型微型计算机系统》2016年第1期48-52,共5页Journal of Chinese Computer Systems

基　　金：山东省优秀中青年科学家科研奖励基金项目(2010BSE14022)资助

摘　　要：推荐系统是解决信息过载问题的有效方法,而协同过滤通过挖掘用户行为信息来预测用户偏好,是现今广泛应用的推荐方法.但传统的协同过滤算法存在数据稀疏,推荐精度不高的问题.而标签信息能够丰富用户(资源)之间的联系,从而提高推荐精度.通过标签信息来构造用户和资源的特征矩阵,进一步融合到基于邻域的协同过滤推荐算法中,预测用户对资源的评分.同时考虑了用户评分的时间上下文影响,降低预测误差.在真实的数据集上验证,该推荐算法与传统协同过滤算法相比,有效的预测用户评分,提高推荐精度.Recommended system is an effective way to solve the problem of information overload. Collaborative filtering, which makespersonalized predictions by mining the behavior information of users, is widely used in recommender systems. But, traditional collabo-rative filtering exist the problems of sparse data and low recommendation accuracy. The label information can enrich the users （ re-source） connection, so as to improve the accuracy of recommendation. The characteristic matrix of the users and resources to be con-structed through the label information, further integration into the collaborative filtering recommendation algorithm based on userneighborhood, prediction of resources score. Considering the time context of user rating, decrease the prediction error. Validation on re-al data sets, the recommendation algorithm and compared to traditional collaborative filtering algorithms, predicting user effectivescore, improve the recommendation accuracy.

关 键 词：协同过滤 标签特征 时间上下文 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]