资源描述：
《基于领域最近邻的协同过滤推荐算法_李聪》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、计算机研究与发展ISSN1000-1239CN11-1777TPJournalofComputerResearchandDevelopment45(9):1532-1538,2008基于领域最近邻的协同过滤推荐算法112李聪梁昌勇马丽1(合肥工业大学管理学院合肥230009)2(西华师范大学商学院四川南充637002)(cnlicong@yahoo.cn)ACollaborativeFilteringRecommendationAlgorithmBasedonDomainNearestNeighbor112L
2、iCong,LiangChangyong,andMaLi1(SchoolofManagement,HefeiUniversityofTechnology,Hefei230009)2(BusinessCollege,ChinaWestNormalUniversity,Nanchong,Sichuan637002)AbstractCurrentlyE-commercerecommendersystemsarebeingusedasanimportantbusinesstoolbyanincreasingnumberofE-comm
3、ercewebsitestohelptheircustomersfindproductstopurchase.CollaborativefilteringisthemostsuccessfulandwidelyusedrecommendationtechnologyinE-commercerecommendersystems.However,traditionalcollaborativefilteringalgorithmfacesseverechallengeofsparseuserratingsandrea-ltimere
4、commendation.Tosolvetheproblems,acollaborativefilteringrecommendationalgorithmbasedondomainnearestneighborisproposed.Theunionofuserratingitemsisusedasthebasisofsimilaritycomputingamongusers,andthenon-targetusersaredifferentiatedintotwotypesthatwithoutrecommendingabil
5、ityandwithrecommendingability.Totheformerusers,usersimilaritywillnotbecomputedforimprovingrea-ltimeperformance;tothelatterusers,domainnearestneighbormethodisproposedandusedtopredictmissingvaluesintheunionofuserratingitemswhentheusershavecommonintersectionsofratingi
6、temclasseswithtargetuser,andthentheneededitemsspaceformissingvaluespredictingcanbereducedtothefewcommonintersections.Thusthesparsitycanbedecreasedandtheaccuracyofsearchingnearestneighborcanbeimproved.Theexperimentalresultsshowthatthenewalgorithmcanefficientlyimprover
7、ecommendationquality.Keywordscollaborativefiltering;recommendationalgorithm;domainnearestneighbor;usersimilarity;MAE摘要协同过滤是目前电子商务推荐系统中广泛应用的最成功的推荐技术,但面临严峻的用户评分数据稀疏性和推荐实时性挑战.针对上述问题,提出了基于领域最近邻的协同过滤推荐算法,以用户评分项并集作为用户相似性计算基础,将并集中的非目标用户区分为无推荐能力和有推荐能力两种类型;对于前一类用户不再计算用户相似性
8、以改善推荐实时性,对于后一类用户则提出领域最近邻方法对并集中的未评分项进行评分预测,从而降低数据稀疏性和提高最近邻寻找准确性.实验结果表明,该算法能有效提高推荐质量.关键词协同过滤;推荐算法;领域最近邻;用户相似性;平均绝对误差中图法分类号TP311收
