粗糙集理论

Roughsettheoryanditsapplications
Zdzis awPawlak
Abstract—Inthispaperrudimentsofthetheorywillbeoutlined,andbasicconceptsofthetheorywillbeillustratedbyasimpletutorialexample,concerningchurnmodelingintelecommunications.Reallifeapplicationsrequiremoread-vancedextensionsofthetheorybutwewillnotdiscusstheseextensionshere.Roughsettheoryhasanoverlapwithmanyothertheoriesdealingwithimperfectknowledge,e.g.,evidencetheory,fuzzysets,Bayesianinferenceandothers.Neverthe-less,thetheorycanberegardedasanindependent,comple-mentary,notcompeting,disciplineinitsownrights.Keywords—roughset,decisionrules,churnmodeling.
1.Introduction
Roughsettheorycanberegardedasanewmathemati-caltoolforimperfectdataanalysis.Thetheoryhasfoundapplicationsinmanydomains,suchasdecisionsupport,engineering,environment,banking,medicineandothers.Thispaperpresentsbasisofthetheorywhichwillbeillus-tratedbyasimpleexampleofchurnmodelingintelecom-munications.
Roughsetphilosophyisfoundedontheassumptionthatwitheveryobjectoftheuniverseofdiscoursesomeinfor-mation(data,knowledge)isassociated.Objectscharac-terizedbythesameinformationareindiscernible(similar)inviewoftheavailableinformationaboutthem.Thein-discernibilityrelationgeneratedinthiswayisthemath-ematicalbasisofroughsettheory.Anysetofallindis-cernible(similar)objectsiscalledanelementaryset,andformsabasicgranule(atom)ofknowledgeabouttheuni-verse.Anyunionofsomeelementarysetsisreferredtoasacrisp(precise)set–otherwisethesetisrough(impre-cise,vague).Eachroughsethasboundary-linecases,i.e.,objectswhichcannotbewithcertaintyclassi ed,byem-ployingtheavailableknowledge,asmembersofthesetoritscomplement.Obviouslyroughsets,incontrasttopre-cisesets,cannotbecharacterizedintermsofinformationabouttheirelements.Withanyroughsetapairofprecisesets,calledthelowerandtheupperapproximationoftheroughset,isassociated.Thelowerapproximationconsistsofallobjectswhichsurelybelongtothesetandtheupperapproximationcontainsallobjectswhichpossiblybelongtotheset.Thedi erencebetweentheupperandthelowerapproximationconstitutestheboundaryregionoftheroughset.Approximationsarefundamentalconceptsofroughsettheory.
Roughsetbaseddataanalysisstartsfromadatatablecalledadecisiontable,columnsofwhicharelabeledbyattributes,rows–byobjectsofinterestandentriesofthetableareat-
tributevalues.Attributesofthedecisiontablearedividedintotwodisjointgroupscalledconditionanddecisionat-tributes,respectively.Eachrowofadecisiontableinducesadecisionrule,whichspeci esdecision(action,results,outcome,etc.)ifsomeconditionsaresatis ed.Ifadeci-sionruleuniquelydeterminesdecisionintermsofcondi-tions–thedecisionruleiscertain.Otherwisethedecisionruleisuncertain.Decisionrulesarecloselyconnectedwithapproximations.Roughlyspeaking,certaindecisionrulesdescribelowerapproximationofdecisionsintermsofconditions,whereasuncertaindecisionrulesrefertotheboundaryregionofdecisions.
Witheverydecisionruletwoconditionalprobabilities,calledthecertaintyandthecoveragecoe cient,areasso-ciated.Thecertaintycoe cientexpressestheconditionalprobabilitythatanobjectbelongstothedecisionclassspec-i edbythedecisionrule,givenitsatis esconditionsoftherule.Thecoveragecoe cientgivestheconditionalproba-bilityofreasonsforagivendecision.
Itturnsoutthatthecertaintyandcoveragecoe cientssat-isfyBayes’theorem.Thatgivesanewlookintotheinter-pretationofBayes’theorem,ando ersanewmethoddatatodrawconclusionsfromdata.
Inthepaperrudimentsofthetheorywillbeoutlined,andbasicconceptsofthetheorywillbeillustratedbyasimpletutorialexampleofchurnmodeling.Reallifeapplicationsrequiremoreadvancedextensionsofthetheorybutwewillnotdiscusstheseextensionsinthispaper.
Roughsettheoryhasanoverlapwithmanyothertheoriesdealingwithimperfectknowledge,e.g.,evidencetheory,fuzzysets,Bayesianinferenceandothers.Nevertheless,thetheorycanberegardedasanindependent,comple-mentary–notcompetingdiscipline,initsownrights.MoreinformationaboutroughsetsandtheirapplicationscanbefoundinthereferencesandtheWeb.
2.Illustrativeexample
Letusstartourconsiderationsfromaverysimpletuto-rialexampleconcerningchurnmodelingintelecommuni-cations,whichisasimpli edversionofanexamplegivenin[1].InTable1,sixfactsconcerningsixclientsegmentsarepresented.
Inthetableconditionattributesdescribingclientpro leare:In–incomingcalls,Out–outgoingcallswithinthesameoperator,Change–outgoingcallstoothermobileoperator,thedecisionattributedescribingtheconsequenceisChurnandNisthenumberofsimilar

cases.
7
Highly Scalable Rough Set Reducts GenerationHighly Scalable Rough Set Reducts Generation...Rough set theory was developed by Zdzislaw Pawlak [1] in the early ...
c Springer-Verlag Berlin Heidelberg 1999 4 Zdzislaw Pawlak called a database. Next, basic operations on sets in rough set theory, the lower and the ...
2014粗糙集方法_数学_自然科学_专业资料。Rough Set Theory Brief Introduction 智能技术课程--YE D Y Introduction ? Developed by Zdzislaw Pawlak inRough...
Indiscernibility Set Approximation Reducts and Core Rough Membership Dependency of Attributes Introduction Rough set theory was developed by Zdzislaw Pawlak in the...
Key words: granular representation calculus; granular rough theory; pure ...(rough set theory)[1?4] 是波兰数学家 Zdzislaw Pawlak 创建的一门独立软...
rough set theory by Pawlak in 1982 [10, 11, 13], many proposals have been made to incorporate probabilistic approaches into the theory [14, 29–31]...
Zhihai WANG, 2012 6 Introduction (4-1) Rough set theory was developed by Zdzislaw Pawlak in the early 1980’s. Representative Publications: ? Z. ...
粗集理论( Rough Set Theory, 简称 RS 理论) 是由 波兰学者 Zdzislaw Pawlak 教授在 1982 年提出的, 它是 一种处理含糊和不确定性问题的新型数学工具, 能有...
ROUGH CONTROL APPLICATION OF ROUGH SET THEORY TO CONTROLROUGH CONTROL APPLICATION...OH 44115 rnunakata@cis.csuohio.edu Zdzislaw PAWLAK Institute of Computer Scie...
1 Introduction Rough set theory, proposed by Pawlak in the early 1980s [4,5], has been widely conceived as an objective mathematical tool to deal ...
Rough Set精华_工学_高等教育_教育专区。Rough Set精华阐述Generated by Foxit PDF...Rough Set 110页 免费 rough set pawlak 51页 免费 A Rough Set Approach t...
热门文档