Factorization Meets the Neighborhood

Factorization Meets the Neighborhood

ID:40070684

大小:281.21 KB

页数:9页

时间:2019-07-19

Factorization Meets the Neighborhood_第1页
Factorization Meets the Neighborhood_第2页
Factorization Meets the Neighborhood_第3页
Factorization Meets the Neighborhood_第4页
Factorization Meets the Neighborhood_第5页
资源描述:

《Factorization Meets the Neighborhood》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、FactorizationMeetstheNeighborhood:aMultifacetedCollaborativeFilteringModelYehudaKorenAT&TLabs–Research180ParkAve,FlorhamPark,NJ07932yehuda@research.att.comABSTRACTprevioustransactionsorproductratings—anddoesnotrequirethecreationofexplicitprofiles.Notably,CFtechniquesrequire

2、nodo-Recommendersystemsprovideuserswithpersonalizedsuggestionsmainknowledgeandavoidtheneedforextensivedatacollection.forproductsorservices.ThesesystemsoftenrelyonCollaborat-Inaddition,relyingdirectlyonuserbehaviorallowsuncoveringingFiltering(CF),wherepasttransactionsareana

3、lyzedinordertocomplexandunexpectedpatternsthatwouldbedifficultorimpos-establishconnectionsbetweenusersandproducts.Thetwomoresibletoprofileusingknowndataattributes.Asaconsequence,CFsuccessfulapproachestoCFarelatentfactormodels,whichdi-attractedmuchofattentioninthepastdecade,r

4、esultinginsignif-rectlyprofilebothusersandproducts,andneighborhoodmodels,icantprogressandbeingadoptedbysomesuccessfulcommercialwhichanalyzesimilaritiesbetweenproductsorusers.Inthisworksystems,includingAmazon[15],TiVoandNetflix.weintroducesomeinnovationstobothapproaches.Thefa

5、ctorandInordertoestablishrecommendations,CFsystemsneedtocom-neighborhoodmodelscannowbesmoothlymerged,therebybuild-parefundamentallydifferentobjects:itemsagainstusers.Thereareingamoreaccuratecombinedmodel.Furtheraccuracyimprove-twoprimaryapproachestofacilitatesuchacompariso

6、n,whichcon-mentsareachievedbyextendingthemodelstoexploitbothexplicitstitutethetwomaindisciplinesofCF:theneighborhoodapproachandimplicitfeedbackbytheusers.Themethodsaretestedontheandlatentfactormodels.Netflixdata.ResultsarebetterthanthosepreviouslypublishedonNeighborhoodmeth

7、odsarecenteredoncomputingtherelation-thatdataset.Inaddition,wesuggestanewevaluationmetric,whichshipsbetweenitemsor,alternatively,betweenusers.Anitem-highlightsthedifferencesamongmethods,basedontheirperfor-orientedapproachevaluatesthepreferenceofausertoanitemmanceatatop-Kre

8、commendationtask.basedonratingsofsimilaritemsbythesameuser.Inasense,CategoriesandSubjectD

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。