一种播存网络环境下的UCL协同过滤推荐方法.pdfVIP

计算机研究与发展 DOI：10．7544／issn1000—1239．2015 JournalofComputerResearchandDevelopment ． 52(2)：475—486，2015 一 种播存网络环境下的UCL协同过滤推荐方法 顾 梁 杨 鹏 罗军舟 (东南大学计算机科学与工程学院 南京 211189) (计算机 网络和信息集成教育部重点实验室 (东南大学) 南京 211189) (guliang@seu．edu．cn) A CollaborativeFilteringRecommendationM ethodforUCL inBroadcast-Storage Network GuLiang，YangPeng，andLuoJunzhou (SchoolofComputerScienceandEngineering，SoutheastUniversity，Nanjing211189) (KeyLaboratory ofComputerNetworkand Inrformation Integration (SoutheastUniversity)，MinistryofEducation， Nanjing211189) Abstract Problemslikebandwidthcongestion，contentredundancyexistinthesharingofinformation resources．Broadcast—Storagenetworkhasaparticularadvantageinsolvingtheseissuesbecauseofits uniquefeatureofone to infinitebyphysicalbroadcast．Uniform contentlabel，UCL，isused to expresstheneedsofusersandhelp usersunderstandtheinformation resourcesinBroadcast—Storage environment．DuetoUCL’Slargequantity，how toguideuserstogettheirpreferredUCLsefficiently isquite significant．Toaddressthisproblem ，thispaperproposesaunifying collaborativefiltering methodwithpopularityandtiming (UCF—PT)forUCL recommendation．First，apairofthresholds aresettoestimatethesparsityofusersandUCLsinthedatasetand determinetheweightsofusers andUCLsin recommendation．ThenUCF—PT predictstheratingsofusersonUCLsbasedon the weightsandgeneratesarecommendation list．Moreover，themethodmakespopularandnew UCLs morelikely to be recommended by considering UCL popularity and using exponentialdecay in recommendation．Experimentsshow that，comparedwith traditionalrecommendationmethods，the methodproposedin thispaperpossessesbetterrecommendation accuracyand ensuresthepopularity andnovelty ofrecommended UCLs． Therefore，itis more suitable for recommending UCLs in Broadcast—Storageenvironment． Key words Broadcast—S