用户属性加权活跃近邻的协同过滤算法  

User-attribute-weighted active K-nearest neighbor's collaborative filtering algorithm

作　　者：王吉源[1] 黎晨[1] 王婵娟[1] Wang Jiyuan Li Chen Wang Chanjuan(School of Information Engineering, Jiangxi University of Science ＆ Technology, Ganzhou Jiangxi 341000, China)

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《计算机应用研究》2016年第12期3625-3629,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(71462018);江西省研究生创新专项基金资助项目(YC2014-S371)

摘　　要：针对现有的基于KNN近邻协同过滤技术,在选择最近邻居时过于依赖评分相似度的问题,提出了一种用户属性加权活跃近邻的协同过滤算法。首先,通过引入用户特征属性并融合最小权重相似度,根据所得的最终相似度生成目标用户的KNN近邻集。然后,从对目标项目已有反馈信息的用户中生成目标项目的活跃用户子群体,并筛选出KNN近邻集中的活跃用户子群体作为目标用户的活跃近邻集,最终产生评分预测。在公开数据集上的实验结果表明,该算法能有效地提高推荐算法的推荐准确度,具有更好的稳定性。Aiming at the problem that the nearest neighbor＇ s collaborative filtering technology based on KNN is extreme dependence on the rating similarity in the choice of nearest neighbors, this paper presented a user-attribute-weighted active Knearest neighbor＇ s collaborative filtering algorithm. First of all, by introducing user＇s feature attributes and fusing minimum weight similarity, the final similarity generated a KNN nearest neighbor set of target users. Users who had feedback from the target items generated active user＇s subpopulations of target items. This paper selected active user＇s subpopulations of KNN nearest neighbors as the active nearest neighbor sets of target users and eventually produced score predicts. Experimental results on the public data sets show that the proposed algorithm can effectively improve the recommendation accuracy and has better stability.

关 键 词：协同过滤 相似度 用户属性 最近邻居集 活跃近邻集 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]