数据流分类中概念漂移和噪声处理研究.pdf

Abstract With ofinformation and ofdata development technologycomputernetwork，lots rapid was innumerous asstock transactions，weather generated fields，suchexchange monitoring， dataareoften network commerceandSOon．These calleddatastreams security,electronical washiddenand becauseof instreamsform．Abundant needtobemined appeared knowledge one branchofdata onthesedatastreams．Asof research urgently primary mining role，and classificationhas valueon datastreamsclassification important applications，SO technology data hasbecomeoneofresearchhot indata streamshavesuch spots mining．However,these and causestraditional characteristics：rapid，continuousnon-repeatability,whichalgorithms werenot whendealwithdatastreams．Ontheone or in hand，information applicable concepts datastreamswill withtimeor theother vary environment，namelyconceptdrifting．On hand， Noisedataisinevitableintherealenvironment．Theclassificationofclassification accuracy noisedata．Andhowtodiscoverand to driftandnoisedata modelaffects adaptconcept by hasbecomea indatastreams effectively big handling