机器学习十大算法：PageRank

《机器学习十大算法：PageRank》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、Chapter6PageRankBingLiuandPhilipS.YuContents6.1Introduction...........................................................1176.2PageRankAlgorithm...................................................1186.3AnExtension:Timed-PageRank....................................

2、...1236.4Summary..............................................................1246.5Exercises..............................................................124References..................................................................1256.1IntroductionLink-bas

3、edrankinghascontributedsignificantlytothesuccessofWebsearch.PageRank[1,7]isperhapsthebestknownlink-basedrankingalgorithm,whichalsopowerstheGooglesearchengine.DuetothehugebusinesssuccessofGoogle,PageRankhasemergedasthedominantlinkanalysismodelontheWeb.ThePageRan

4、kalgorithmwasfirstintroducedbySergeyBrinandLarryPageattheSeventhInternationalWorldWideWebConference(WWW7)inApril1998,withtheaimoftacklingsomemajordifficultieswiththecontent-basedrankingalgorithmsofearlysearchengines.Theseearlysearchenginesessentiallyretrievedrel

5、evantpagesfortheuserbasedoncontentsimilaritiesoftheuserqueryandtheindexedpagesofthesearchengines.Theretrievalandrankingalgorithmsweresimplydirectimplementationofthosefrominformationretrieval.However,startingfrom1996,itbecameclearthatthecontentsimilarityalonewa

6、snolongersufficientforsearchduetotwomainreasons.First,thenumberofWebpagesgrewrapidlyduringthemiddletolate1990s.Givenanyquery,thenumberofrelevantpagescanbehuge.Forexample,giventhesearchquery“classificationtechnique,”theGooglesearchengineestimatesthatthereareabout

7、10millionrelevantpages.Thisabundanceofinformationcausesamajorproblemforranking,thatis,howtochooseonly10to30pagesandrankthemsuitablytopresenttotheuser.Second,contentsimilaritymethodsareeasilyspammed.Apageownercanrepeatsomeimportantwordsandaddmanyremotelyrelated

8、wordsinhis/herpagestoboosttherankingsofthepagesand/ortomakethepagesrelevanttoalargenumberofpossiblequeries.117©2009byTaylor&FrancisGroup,LLC118PageRankFromaround1996,researchersina