ch04数据流挖掘2研讨.pptx

下载文档 降价啦

7
0
约 46页
2016-12-06 发布于湖北
举报
版权申诉
保障服务

ch04数据流挖掘2研讨.pptx

1、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

ch04数据流挖掘2研讨

MiningDataStreams(Part2)MiningofMassiveDatasetsJureLeskovec,AnandRajaraman,JeffUllmanStanfordUniversityNotetootherteachersandusersoftheseslides:Wewouldbedelightedifyoufoundthisourmaterialusefulingivingyourownlectures.Feelfreetousetheseslidesverbatim,ortomodifythemtofityourownneeds.Ifyoumakeuseofasignificantportionoftheseslidesinyourownlecture,pleaseincludethismessage,oralinktoourwebsite:Today’sLectureMorealgorithmsforstreams:(1)Filteringadatastream:BloomfiltersSelectelementswithpropertyxfromstream(2)Countingdistinctelements:Flajolet-MartinNumberofdistinctelementsinthelastkelementsofthestream(3)Estimatingmoments:AMSmethodEstimatestd.dev.oflastkelements(4)CountingfrequentitemsJ.Leskovec,A.Rajaraman,J.Ullman:MiningofMassiveDatasets,2(1)FilteringDataStreamsFilteringDataStreamsEachelementofdatastreamisatupleGivenalistofkeysSDeterminewhichtuplesofstreamareinSObvioussolution:HashtableButsupposewedonothaveenoughmemorytostoreallofSinahashtableE.g.,wemightbeprocessingmillionsoffiltersonthesamestream4J.Leskovec,A.Rajaraman,J.Ullman:MiningofMassiveDatasets,ApplicationsExample:EmailspamfilteringWeknow1billion“good”emailaddressesIfanemailcomesfromoneofthese,itisNOTspamPublish-subscribesystemsYouarecollectinglotsofmessages(newsarticles)PeopleexpressinterestincertainsetsofkeywordsDeterminewhethereachmessagematchesuser’sinterest5J.Leskovec,A.Rajaraman,J.Ullman:MiningofMassiveDatasets,FirstCutSolution(1)GivenasetofkeysSthatwewanttofilterCreateabitarrayBofnbits,initiallyall0sChooseahashfunctionhwithrange[0,n)HasheachmemberofsStooneofnbuckets,andsetthatbitto1,i.e.,B[h(s)]=1Hasheachelementaofthestreamandoutputonlythosethathashtobitthatwassetto1OutputaifB[h(a)]==16J.Leskovec,A.Rajaraman,J.Ullman:MiningofMassiveDatasets,FirstCutSolution(2)CreatesfalsepositivesbutnofalsenegativesIftheitemisinSwesurelyoutputit,ifnotwemaystilloutputit7FilterItem0010001011000HashfunchBitarrayBJ.Leskovec,A.Rajaraman,J.Ullman:MiningofMassiveDatasets,FirstCutSolution(3)|S|=1billionemailaddresses|B|=1GB=8