网页排序中的随机模型及算法

《网页排序中的随机模型及算法》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、北京交通大学博士学位论文网页排序中的随机模型及算法姓名：刘玉婷申请学位级别：博士专业：运筹学与控制论指导教师：马志明20090401北京交通大学博+学位论文中文摘要成果，也验证了这一想法，为以后更多的理解排序问题奠定了基础．关键词：信息检索；排序问题；静态排序；动态排序；排序联合问题；马氏骨架过程；BrowseRank算法；监督学习分类号：0211北京交通大学博士学位论文ABSTRACTAstheWorld、MdeWebgrowsrapidly,searchengineshavebecomethemostpop-u

2、lartoolstoaccessthelargevolumeofinformationfromit．AndthekeyfactorofthesearchengineistherankingmodelofWebpages，whichcontainstWOtypes：staticrankanddynamicrank．Inpast，differentapproacheshavebeendesignedtOsolvethesetwoproblemsseparately．Inthisthesis，weanalyzethemo

3、nthesamebase—stochasticprocessmodel，anddesignnewalgorithmstosolvethemefficientlyandeffectively．Firstly,weestablishaframeworkonMarkovskeletonprocesstocomputethepageimportancebyinvestigatingrealbrowsingbehaviorsofusers．Wefindthattheimpor-tanceofWebpagesonusers’v

4、iewisdeterminedbytwofactors：visitingfrequencyandthelengthofthestayingtimeonthem．AndweprovethatthestationarydistributionoftheMarkovskeletonprocess(itexistsunderspecialconditions)cantakethesetwofac—torstogetherintotheaccount．Withinthisframework，wedesignagroupofe

5、ightnewnovelalgorithmsallreferredtOasBrowseRank，tocomputethepageimportancebasedonthecontinuous—timetime-homogeneousMarkovprocess，whichisoneofthreespe—cialcasesoftheMarkovskeletonprocess．Andfromtheexperimentalresults，wefindBrowseRankoutperformsotherbaselinealgo

6、rithms，suchasPageRankandTrustRank．Itistestifiedthatstochasticprocessmodelismoreeffectiveandefficientforpageim—portancecalculation．Moreimportant，ThefirstoneoftheseeightalgorithmsWaspre—sentedinSIGIR’08，andnowithasbecomethemostpopularalgorithminthefieldofcommerc

7、ialsearchengines．Suchinnovatoryalgorithmwasevaluatedbyvariousre—portagesasthenextPageRankandwillimpacttheprogressofsearchengineshighly,suchas’’Microsofttriestoone-upGooglePageRank’’byCNET．com．’’BrowseRank：TheNextPageRank，SaysMicrosoft”byWebProNewsandSOon．Secon

8、dly,webuildasupervisedframeworkforrankaggregationbasedonMarkovChain．Withinthisframework，Wenotonlygeneralizesomeunsupervisedalgorithmstosupervisedones，butalsodesignanewapproachrefer