基于圖的個性化音樂推薦.pdf

1、ADissertationSubmittedtoZhejiangUniversityfortheDegreeofMasterofEngineering⑧TITLE：魚!壘乜h二旦壘墨金魚里殳!墨Q壘壘!至圣殳魚MusicRecommendationAuthor：Supervisor：￡!Q￡皇璺曼Q!g塾墜旦gh皇旦SubmittedDate：II壘旦：星Q!圣AbstractAstheinformationexplosioneraap

2、proaches，theresourcetypesandthewavstetrieVelnformationhavewidelychangedIntemetmusicradio，asabmdnewwaVtormusicentertainmenthasbecomemoreandmorepopularbecauseofitsfriendlvuserInterfaceandpromptlymusicupdateConsideringtheev

3、olutionarytrendsforthewldelYusefortheintemetradio，thisthesisfocusesonthemusicrecommendersvstemfor1nIernetradiomodelWeproposetwoalgorithmstosolvethepersonalizedmusicrecommendationproblemfordifferentsizeofdataf’1rstlytheth

4、esisintroducesanovelpersonalizedalgorithmbasedontagandrating(abbrPRABOTRalgorithm)Weabstractthemusic腳ioasamodelconsistingjtypesofobjects(usertagandmusic)Afterconstructingtherelationm枷ces鋤ongdlffe陀n‘typesofobjects，thealgo

5、rithmusessubspacelearningtechniquetomapeacho％ectmtoanotherlowIdimensionspaceWiththenewcoordinateofeachobjectinthelearningsubspaee，PRABOTRalgorithmusesKnearestneighbormethodforeaChuserandrecommendsthenearestsongstothemsec

6、ondlyafterremovingthetagobjectsinPRABOTRmodel，thethesisDroposes粕otherstate。ofartalgorithmbasedonratingforlargescaledata(abbralgorithm)，whichhasbeenintegratedintothe9skyrecommendercommercialuseIt’sa‘cash’propagationmodelc

7、onsisting4layersalgorithmmakesthetargetuserLS唯RsystemforLS睜Rasthestartingpoint，andspreads‘cash，betweenadjacen，，、tlayerStogettheresulting‘cash’valueforeverysongInthefinalstep，LS’蝴Rrecommendshighlyratedsongstotheta唱etusera

8、ccordingtothecash，valueThirdlythethesisUSeStwocriterions(precisionandnomlalizeddiscountedcumulativegain)tocomparetheproposedalgorithmswithseveralclassicalrecommendationalgorithmsAndtheexperimentresultshowsthattheproposed