个性化微博信息流推荐技术-研究

《个性化微博信息流推荐技术-研究》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、个性化微博信息流推荐技术研究关键词：微博推荐；信息检索；主题模型；协同过滤；个性化；冷启动论文类型：应用研究-II-万方数据兰州交通大学硕士学位论文AbstractMillionsofpeopleturntomicrobloggingservicestogatherreal-timenewsoropinionaboutpeople,things,oreventsofinterest,becauseitisveryconvenientandfastforusers.Asmicroblogginggrowsinpopularity

2、inrecentyears,microbloggingservicesarebecomingthekeyroleofsocialnetworkstosupportinformationsharing.However,moreandmoreuserstodayarefacingthechallengeofinformationoverloading.So,howtorecommendtheinterestingandimportanttweetsforuseristhekeyissue.Themainchallengesofth

3、epersonalizedtweetsstreamrecommendationisthecontradictionbetweenthesmallsizeofthemicroblogmessagesandthesocialfeaturesofmicroblogsandthereforehowtodetectpersonalinterestexactlyisthekeyprobleminthisthesis.Mostofthecurrentresearchesonthepersonalizedtweetsstreamrecomme

4、ndationperformbadlybecauseofthelimitationsofmodelsresearchandthechallengesofmicroblogsdataitself.Inthisthesis,weintegratetheTF-IDF(termfrequency–inversedocumentfrequency)modelsofthesimilaritybetweenstreamsoftweetsandthesimilarityscoresbetweenstreamsoftweetsbasedonLD

5、A(LatentDirichletAllocation)modelstosolvethisproblem.Theresearchcontentsareasfollows:Firstly,anapproachbasedonTF-IDFisproposed.Inordertocapturepersonalinterests,theapproachimprovestheTF-IDFmodelbycombiningtheweightofsingletermsandpairsoftermsandevaluatesthesimilarit

6、ybetweenthesetofuser’stweetsandthestreamoftweetscomingtousersbasedontheideaofcollaborativefiltering.Moreover,themodelstudiesthecold-startproblemsandpersonalfeaturesofuserstooptimizethequeueofthetweetsreceivedbyusers.Secondly,thisthesisalsoborrowsthemachineryoflatent

7、variabletopicmodelslikethepopularunsupervisedmodelLDAwhichhavebeenappliedwidelytoproblemsintextmodeling.Thesemodelsdistillcollectionsoftextdocuments(tweets)intodistributionsofwordsthattendtoco-occurinsimilardocuments.Sotheapproachusestopicstocomputethesimilarityscor

8、esbetweenstreamsoftweets.Finally,thisthesiscombinestheimprovedmodelsbasedonTF-IDFandLDAmodelsforpersonalizedtweetsstreamrecommendation,and