推荐系统中分布式混合协同过滤方法.PDFVIP

2016年4月 北 京 邮 电 大 学 学 报 Apr.2016 第39卷 第2期 Journal of Beijing University of Posts and Telecommunications Vol.39 No.2 摇 摇 文章编号:1007鄄5321(2016)02鄄0025鄄05 DOI:10.13190/ j.jbupt.2016.02.005 推荐系统中分布式混合协同过滤方法 王晓军 (南京邮电大学 信息网络技术研究所,南京210003) 摘要:传统协同过滤方法面临数据稀疏问题,稀疏的用户-项目关联数据将产生不准确的相似用户或项目,为了改 善推荐质量,提出一种基于Map Reduce 的混合协同过滤方法. 该方法利用用户特征和用户-项目评分数据构造项 目偏好向量,然后使用模糊K鄄Means算法对项目进行聚类,并从每个项目簇中选择相似项目,最后组合所有项目簇 的预测结果作出推荐. 实验结果显示,该方法能缓解数据稀疏问题,改善推荐精度. 关摇 键摇 词:分布式框架;个性化推荐;协同过滤;模糊聚类 中图分类号:TP391摇 摇 摇 摇 文献标志码:A A Distributed Hybrid Collaborative Filtering Method in Recommender Systems WANG Xiao鄄jun (Institute of Information and Network Technology,Nanjing University of Posts and Telecommunications,Nanjing210003,China) Abstract:Addressingtheinformationoverloadingproblem,thecollaborativefiltering isan effectivetech鄄 nique,andextensively appliedinrecommender systems. Itmakepredictionsbyfindinguserswith similar taste or items that have been similarly chosen. However,as the number of users or items growsrapidly, the traditionalcollaborativefilteringapproachissufferingfromthedatasparsityproblem. Thesparseuser鄄 item associationscangenerateinaccurateneighborhoodforeachuseroritem. Adistributedhybridcollab鄄 orative filtering method was proposed based on Map Reduce,aiming at improving the recommendation quality. This methodutilizesuserfeaturesandratingstoconstructitempreferencevectors. Then,itclus鄄 tersitemsusingfuzzyK鄄Meansalgorithm,andrespectivelychoosessimilaritemsfromeachclustering ,fi鄄 nally it combinesall predictionsfrom each clustering and makesrecommendation. Experiments show that the distributed hybrid collaborative filtering method can help reduce the sparsity probl