CorrespondingauthorTel3458243199fax3458243317emailopcdecsaiugres1EmailAGonzalezdecsaiugres2Emailviladecsaiugres0888613X99 ID: 105647
Download Pdf The PPT/PDF document "Dealingwithuncertaintyandimprecisionbyme..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
DealingwithuncertaintyandimprecisionbymeansoffuzzynumbersA.Gonzalez,O.Pons,M.A.VilaDepto.CienciasdelaComputacineI.A.,E.T.S.deIngenieriaInformatica,UniversidaddeGranada,18071Granada,SpainReceived1July1998;accepted1February1999Theproblemofthecombinationofimprecisionanduncertaintycombinationfromtheapproximatereasoningpointofviewisaddressed.Animpreciseanduncertainin-formationcanberepresentedasafuzzyquantitytogetherwithacertaintyvalue.Inordertosimplifytheuseofsuchinformation,itisnecessarytocombinetheimprecisionanduncertaintyofthefuzzynumber.Inthispaperweproposeamethodforcombiningthembasedontheuseofinformationmeasures.The®rststepconsistsintruncatingthefuzzynumberbythecertaintyvalue.Sincenon-normalizedfuzzynumbersarediculttouse,wetransformthetruncatedfuzzynumberintoanormalizedfuzzynumberwhichcontainsthesameamountofinformation.Toformalizethisprocess,wedevelopatheoreticalcontextfortheinformationmeasuresonfuzzyvalues.Westudythefuzzynumberstransformationanditsproperties,andgiveanapproximatereasoninginter-pretationtotheapproach.1999ElsevierScienceInc.Allrightsreserved.Uncertainty;Imprecision;Fuzzynumber;Implicationfunction;InformationInternationalJournalofApproximateReasoning21(1999)233±256 *Correspondingauthor.Tel.:+34-58-243199;fax:+34-58-243317;e-mail:opc@decsai.ugr.es1E-mail:A.Gonzalez@decsai.ugr.es2E-mail:vila@decsai.ugr.es0888-613X/99/$±seefrontmatterÓ1999ElsevierScienceInc.Allrightsreserved.PII:S0888-613X(99)00024-9 IntroductionTheco-existenceofimprecisionanduncertaintywithinaconcretedatumappearsinmanyapplications.Forexample,inthestudyofoptimizationmethodsinfuzzygraphs[4]orintheframeworkofuncertainfuzzydatabases[1].Inthesetwocases,thegivensolutionsgiverisetoaseriesofinconveniencesderived,mainly,fromtheuseofnon-normalizedornon-trapezoidalfuzzysets,respectively.Inthispaperweinvestigatetheproblemsassociatedwiththecombina-tionofimprecisionanduncertaintyfromtheapproximatereasoningpointofview.Todothat,weusethegeneraltransformationfunctionTintro-ducedin[14]thatwillallowustorelateourresultstootherapproachesoftheliteratureandthatwillopennewwaysforthetreatmentofthisprob-lem.Thistypeofinformationcanbeexpressed,ingeneral,byanimprecisevalueA(represented,forexample,byatrapezoidalfuzzynumber)togetherwithacertaintylevelaassociatedwithsuchvalue.Thesituationcanbeformulatedasaconditionalexpressioninthefollowingterms.IfthedatumistotallytruethenitsvalueisA.Sincewehaveacertaintylevela1,thegeneralizedmodusponenscouldbeformulatedas:Ifthecertaintylevelis1,thenthevalueisA.Ifthecertaintylevelisa1,thenthevalueisA0:Thissituationisequivalenttothegenericcase: Therefore,anaturalwaytosolvetheproblemistoconsiderthatthedatumwearehandlingisA0de®nedas:lA0 xI a;lA xwhereIisamaterialim-plicationfunctionwhichre¯ectstheinterpretationgiventothecompatibilitydegree.Thereexistintheliteraturetwomainwaysofdealingwithimpreciseanduncertaindataandcanbeinterpretedasfollows:1.ToTruncate:Ifthedatumis a;A,thenA0isde®nedbythemembershipfunctionlA0 xmin a;lA xwhichdirectlyimpliesthatweareusingMamdani'simplicationinourreasoning.2.ToExpand:Ifweassumethataisanecessity,thenA0isgivenbythemem-bershipfunctionlA0 xmax 1ÿa;lA x,whichcorrespondstoKleene±Dienes'implicationasfoundationofourreasoning.Thesetwoapproachescorrespondtothedisjunctiveandconjunctiverepresentationoftheinferencerule,respectively.234A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 ourpointofview,weunderstandthattheuseoftheseimplicationfunctions[2,3,15]forinformationrepresentation(whichisour®nalobjective)couldinduceanerror,sincedatumA0weareusingwillbeevaluatedintermsofcompatibilitywithotherdataand,inthesecases:1.Mamdani'simplicationresultsinadecreaseofthecompatibilitytolevelainanycase.Thisresultseemstobereasonableasaisthecertaintydegreebutitobligesustoworkwithnon-normalizedfuzzynumbers.2.Kleene±Dienes'implicationimposesthatanydatumwouldbecompatiblewithA0atleastatlevel1ÿa.Thisresultmaynotbesuitableforsomeapplications,sinceitassignsthesamepossibilitytoallthepointsoftheunderlyingdomainindependentlyfromthedistancetothesupportsetofA.Letusthinkthat,forexample,ifAisveryheavywithcertaintya,theval-uescloseto0willhavethesamepossibilitythanthoseclosetothesupportofveryheavy.Ourmainobjectiveisto®ndatransformationfunctionthat,basedondierentcriteria,ensuresusasuitablechange.Theintuitiveideasusedto®ndsuchatransformationfunctionare:·TotruncateAatlevela(weobtainAa).·TonormalizeAa(weobtainAT).Ifweassumethetranslationofuncertaintyintoimprecision,thenimpreci-sionofATmustbelargerthanAimprecisionbutATwillneverbede®nedonthewholedomain.TheideaistoincreasesuchimprecisionaroundthesupportsetofvalueA.ThetransformationusedisaccordingtoequitativedistributionofimprecisiononthesupportofAwhichisvalidwhennomoreinformationisprovided,i.e.,imprecisionisdistributedaccordingtoametricwhichtakesintoaccountthenearnesstotheoriginalinformation.Followingtheseideas,whenwehavetheinformationthatXisblackwithcertaintya,wewillnevergiveapositivepossibilitytocolourwhitebuttocoloursnearenoughtoblackde-pendingonvaluea.ThiswayofreasoninghasnotbeenusedyetandwillpermitustoensurethattheinformationamountprovidedbyanuncertainimprecisevalueAisthesameastheinformationprovidedbyitstransformationAT,whichisfullytrueandnormalized.Thepaperisorganizedasfollows.InSection2,thepreliminaryconceptsandthenotationusedareintroduced.InSection3,theaxiomaticde®nitionofaninformationfunctiononfuzzynumbersanditspropertiesaregiven.Basedonthisinformationfunction,atransformationforfuzzynumbersisintroducedinSection4.Thistransformationensuresthattheamountofinformationbeforeandafteritsapplicationremainsequal.InSection5,wearegoingtoprovethatthetransformationfunctionde®nedisanimplicationfunction.Todoit,wearecheckingthatalltheconditionsanimplicationfunctionmusthold,arealsoheldbyourtransformationfunction.Finally,inSection6,themainconclusionsofthisworkaresummarized.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256235 PreliminaryconceptsAfuzzyvalueisafuzzyrepresentationabouttherealvalueofaproperty(attribute)whenitisnotpreciselyknown.Inthispaper,accordingtoGoguen'sFuzzi®cationPrinciple[10],wewillcalleveryfuzzysetofthereallinefuzzyquantity.Afuzzynumberisaparticularcaseofafuzzyquantitywiththefollowingproperties.De®nition1.ThefuzzyquantityAwithmembershipfunctionlA xisafuzzynumberi:1.8a20;1;Aafx2RjlA xPag(a-cutsofA)isaconvexset.2.lA xisanupper-semicontinuousfunction.3.ThesupportsetofASupp Afx2RjlA x0gisaboundedsetofR,whereRisthesetofrealnumbers.Thegivende®nitionisbasedonthede®nitiongivenbyDuboisandPrade[7]butwedonotrequireeithernormalizationorthatthemodalintervalisasingleton.Wewilluse~Rtodenotethesetoffuzzynumbers,andh AtodenotetheheightofthefuzzynumberA.Forthesakeofsimplicity,wewillusecapitallettersatthebeginningofthealphabettorepresentfuzzynumbers.Theintervalaa;ba(seeFig.1)iscalledthea-cutofA.Sothen,fuzzynumbersarefuzzyquantitieswhosea-cutsareclosedandboundedintervals:Aaaa;bawitha2 0;1.ThesetSupp Afx2RjlA x0giscalledthesupportsetofA.Ifthereis,atleast,onepointxverifyinglA x1wesaythatAisanor-malizedfuzzynumber.Sometimes,atrapezoidalshapeisusedtorepresentfuzzynumbers.Thisrepresentationisveryusefulasthefuzzynumberiscompletelycharacterizedby Fig.1.Fuzzynumber.236A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 parameters m1;m2;a;bandtheheighth AasshowninFig.2.Otherparametricalrepresentationforfuzzynumberscanbefoundin[5].Wewillcallmodalsetallvaluesintheintervalm1;m2,i.e.,thesetfx2Supp Aj8y2R;lA xPlA yg.Thevaluesaandbarecalledleftandrightspreads,respectively.Whenafuzzynumberisnotnormalized,thissituationcanbeinterpretedasalackofcon®denceintheinformationprovidedbysuchnumbers[6,11].Infact,theheightofthefuzzynumbercouldbeconsideredasacertaintydegreeoftherepresentedvalue.Ontheotherhand,ifweassumetheseconsiderations,normalizedfuzzynumbersrepresentimprecisequantitiesonwhichwehavecompletecertainty.Aswewillseealongthispaper,thisuncertaintycanbetranslated,usingsomesuitabletransformations,intoimprecision,takingintoaccountthatthelesstheuncertainty(orthemorethecertainty)aboutafuzzynumber,themoreistheimprecisionofsuchanumber.Thistransformationwillbedoneinsuchawaythattheamountofinformationofthefuzzynumberwillbeconstantbeforeandafterthemodi®cation.3.AninformationmeasureonfuzzyvaluesAspointedoutinSection2,wearegoingtotranslateuncertaintyintoim-precisionandviceversaundercertainconditions.Themostimportantoftheseconditionsisthattheamountofinformationprovidedbythefuzzynumberremainsequalbeforeandafterthetransformation.Sothen,the®rststepistode®neaninformationfunctionforfuzzynumbers.Weproposeanaxiomaticde®nitionofinformation,partiallyinspiredinthetheoryofgeneralizedinformationgivenbyKampedeFeriet[13]andthatcanberelatedtotheprecisionindexes[8]andthespeci®cityconcept,introducedbyYagerin[16]. Fig.2.Trapezoidalfuzzynumber.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256237 on2.LetD~RjRD;wesaythattheapplicationIde®nedasI:D!0;1isaninformationonDifitveri®es:1.I A18A2R,2.8A;B2Djh Ah BandAB)I B6I A:The®rstconditionmeansthatrealnumbersaretotallyinformativeand,thesecondone,thatconsideringtwofuzzynumberswiththesameheight,ifoneofthemiscontainedintheotherone,thenitisobviousthatthe®rstone,whichismoreprecise,isalsomoreinformative.Thegivende®nitionofinformationisverysimilartothede®nitionoftheprecisionindex,infact,whenappliedtonormalizedfuzzynumbers,bothofthemcoincide.Thiscoincidenceisveryreasonablebecause,whenthereisnouncertainty,informationisequivalenttoprecision.Inthisway,theinforma-tionfunctionisageneralizationoftheprecisionindexes.De®nition3.LetA2Dbeafuzzynumber.WesaythatAhasthemaximuminformationorthatAistotallyinformativewithrespecttoIiI A1.Obviously,realnumbersaretotallyinformativewithrespecttoanyinfor-mationmeasureI,thatis,8r2R;I r1.Theinformationaboutfuzzynumbersdependsondierentfactors,inparticular,onimprecisionandcertainty.Wefocusongeneraltypesofinfor-mationrelatedonlytothistwofactors.Tocomputeameasureoftheimprecisioncontainedinafuzzynumber,wewillconsiderameasureoftheimprecisionofitsa-cuts,whichareclosedin-tervalsonwhichthefollowingfunctionisde®ned:8A2~R;fA abaÿaaifa6h A;0otherwise:(Fromthisimprecisionfunctiononthea-cuts,wede®nethetotalimprecisionofafuzzyvalueasacombinationoftheimprecisionineverylevela.Whena0,wewillconsiderthatfA 0isthelengthofthesupportset.De®nition4.Theimprecisionofafuzzynumberisde®nedasfollows:f:~R!R0;8A2~R;f AZh0 AfA ada:238A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 imprecisionfunctionfcoincideswiththeareabelowthemembershipfunctionofthefuzzyvalue,asshowninFig.3.Obviously,itisheldthat8A;B2~RjAB)fA a6fB a8a20;hwhenh Ah Bhand,thereforef A6f B:Relatedtotheheight(certainty)andtheimprecisionofafuzzyvalue,wede®nethefollowinggeneraltypeoffunctionon~R:IF:~R!0;1;IF AF h A;f A:Thefollowingresultguaranteesthat,forcertaintypesofFfunctions,IFisainformationfunctiononR.Proposition1.LetF: 0;1R0!0;1suchthat1.F 1;01;2.8y;z2R0jy6z)F x;z6F x;y8x2 0;1:then,IFisaninformationfunctionon~R.Proof.LetA2R.Thenobviouslyh A1andf A0.Then,IF AF 1;01.LetA;B2~RjABandh Ah B.Thenf A6f Banditisveri®edthatF h B;f BIF B6F h A;f AIF Aandtherefore,IFisaninformationon~R.WhenFveri®esthepreviousconditions,wewillcallfunctionIFanF-information.Inthisway,associatedwithaclassoffunctions,wecanbuildsomeparticulartypesofinformationon~Rwiththepropertyofnotdepending Fig.3.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256239 thepositionthefuzzyvaluehasonR,asshowninthefollowingproposi-tion.Proposition2.LetFbeafunctionverifyingconditionsestablishedinProposi-tion1,A2~Randt2R.Then,IF AIF At:Proof.IfA2~Randt2RthenAt2~RandlAt zlA zÿt8z2R.Be-sides,h Atsupz2RflAt zgsupz2RflA zÿtgsupz2RflA zgh Aand AtaAat,resultingthatfA afAt aandf Af At;thereforetheresultisimmediate.Therearemanywaystobuildinformationfunctionsbut,forourpurpose,wearede®ninginformationassociatedwithaparticularfunction.ThisF-informationwillpermit,subsequently,thede®nitionoftransformationsthatkeepconstanttheamountofinformationafuzzynumberprovides.LetusconsiderthefunctionF: 0;1R0!0;1;F x;y xky1;k2R;thattriviallyveri®estheconditionsestablishedinProposition1.Hence,wecande®nethefollowingF-information.De®nition5.Wede®nethefunctionIF:~R!0;1;8A2~R;IF A h Akf A1;whereh Aisthefuzzynumberheight,f AistheimprecisionassociatedwithAandk60aparameterwhichdependsonthedomainscale(inSection5,thisparameteriswidelyexplained).Evidently,byProposition1,IFisaninformationfunctionand,ittriviallyfollowsthat8A2~R;06IF A6h A61:Ascanbeimmediatelydeducedfromitsde®nition,informationIFisalwaysboundedbythefuzzynumberheight.Therefore,fuzzynumberswithmaximuminformationwithrespecttoIFmustalsohavemaximumheight(h A1)and,consequently,minimumimprecision(f A0).240A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 fuzzynumbersshowninFigs.4(a)and(b)providethesameinforma-tionas 0:511 1310:25assumingk1.ThefuzzynumbersshowninFigs.4(b)and(d)arethesamefuzzynumbersexpressedindierentdomainscales.Astheinformationprovidedbybothnumbersshouldbethesame,thekparametermustbeadaptedtothescalechangesconsideringabaseorreferencescalewherekissetto1(inthiscasekilometers) 131 1 1100030001;sothen,ifthebasescaleiskilometersandthecurrentscaleinmeters,kpa-rametermustbesetto1=1000. Fig.4.Fuzzyvaluesexamples.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256241 caseofFig.4(c)isnotreallyafuzzynumberrepresentationifwestrictlyfollowDe®nition1,butitisveryillustrativetoseehowthesetypesoffuzzyquantitieswithin®nitesupport,provideinformation0,asf A1.Wecouldalsowonderwhichintervalhasthesameinformationamountasaconcretenormalizedtrapezoidalfuzzynumber.Theanswertothisquestionisthefollowing±LetussupposeournormalizedtrapezoidalfuzzynumberisthegeneralonerepresentedinFig.5(a).Oneofthepossibleintervals(theonecenteredinthemodalset)withthesameinformationasthefuzzynumberisrepresentedinFig.5(b)expressedbyBm1ÿ a2;m2 b2.ItcanbeeasilyprovedthatI AI BandthatBE A,i.e.themeanvalueofAinthesenseofDuboisandPrade[9,12].Oncewehaveaninformationfunctiononfuzzynumbers,wewanttouseittode®netransformationswhichpreservesuchinformationfunctionvalue.Theideaisto®ndanequivalentrepresentationoftheconsideredfuzzynumberinsuchawaythatwechangeuncertaintybyimprecisionkeepingconstanttherelationshipbetweenthemde®nedbytheinformationfunction.4.Fuzzynumberstransformations4.1.BasicmodelTheaimofthetransformationsweareproposinginthissectionis,basically,tobeabletomodifytheheightofafuzzynumberbutkeepingtheinformationcontainedinit.Thereasonfordoingthisisthat,inmostapplications,itisveryconvenientthatfuzzynumbersarenormalized(simplicity,betterunderstand-ingforusers,etc...).Givenafuzzynumber,atransformationonitwillgiveanotherfuzzynumberwiththesameinformationamountbutdierentheight.Sothen,tode®netransformations,wewillrequestthattheinformation Fig.5.Normalizedtrapezoidalfuzzynumberandthecorrespondinginterval.242A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 remains®xed,i.e.,wewillmodifycertaintyandimprecisionbutkeepingconstanttherelationbetweenbothnumbers,whichisde®nedbytheinformationfunction.Thede®nitionoftransformationwillbeobtainedfromtheconditionofequalityintheinformationbut,asa®rststep,wemustestablishwhatweunderstandfortransformationofafuzzynumberonasubsetof~R.De®nition6.Letusconsidera2 0;1andtheclassoffuzzynumbersD~R.WesaythatTa:D!~RisatransformationforaninformationfunctionIonD,ifitveri®esthat:1.Ta A2D;2.h Ta Aa;3.I Ta AI A8A2D:Inthisway,foraheightlevela,Ta Aneednotexistbut,ifitdoes,itmustverifytheconditionsabove.De®nition7.GiventhetransformationTa,wesaythatA2Distransformablefora 0;1ifthereexistsTa A.WewilldenoteH Afa2 0;1j9Ta Agthesetoflevels,whereAistransformable.Thoughmostoftheresultsobtainedherecanbegeneralizedforanytypeoffuzzynumber,wewillfocusontrapezoidalonesforthesakeofsimplicityinthetransformationfunction.WewillnotebystheclassoftrapezoidalfuzzynumbersonR.GivenafuzzynumberA2s,wearelookingfortheconditionsthatanotherfuzzynumberB,with®xedheighta2 0;1,mustholdtohavethesamein-formationamountasA.Proposition3.LetA;B2sbetwofuzzynumberswithheightsh AaAandh BaB,respectively.Then,IF AIF B()fB 0fB aBfA 0fA aA 2kD aA;aB;whereD aA;aB aBÿaAaAaB:Proof.Itisimmediatefortrapezoidalfuzzynumbers,takingintoaccountthatf A fA 0fA aA=2aA:A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256243 is,thesumofbaseimprecisionandmodalimprecisionmustbemod-i®edbythevalue 2=kD aA;aBforAcanbetransformedintoafuzzynumberBof®xedheight.Besides,ifwepretendtoputuptheheightofA aAaB,thenD aA;aBispositiveandthesumofbaseimprecisionandmodalimprecisionofBmustaugment;ontheotherhand,toputdowntheheight aBaA,sinceD aA;aBisnegative,imprecisionmustbedecreased.Whentheheightis®xed,itisobviousthatimprecisionremainsequal.Sothen,therelationbetweenuncertaintyandimprecisionisthefollowing:·Anincreaseofcertaintymeansanincreaseofimprecision.·Adecreaseofimprecisionmeansadecreaseofcertainty.Proposition3permitsustode®neatransformationassumingthat:1.Modalimprecisionispreserved.2.Theincrease/decreaseofimprecisionisequallydistributedintherightandleftsidesofthefuzzynumberindependentlyfromitsshape.De®nition8.LetA2sbeafuzzynumbersuchthatAf m1;m2;a;b;aAg;wherem1;m2;aandbareshowninFig.2andaAistheheightofA.Leta2 0;1.WewilldenoteD aA;aDandde®neTa Am1;m2;a Dk;b Dk;aforthoseainwhichthetransformationmakessense.Proposition4.Taisatransformationfortrapezoidalfuzzynumbers.Proof.LetusassumethatthereexistsTa Afora2 0;1.Then,obviouslyTa A2sandh Ta Aa.Ontheotherhand,fTa A 0fTa A afA 0fA aA 2kD:ByProposition3,IF AIF Ta AandusingDe®nition6,Taisanin-formationons:De®nition9.LetAf m1;m2;a;b;aAgbeatrapezoidalfuzzynumber.Wede®nethelowestlimitofthetransformationasl Amax aAkaaA1; aAkbaA1:Itcanbeprovedimmediatelythatl Aisanumberintheinterval 0;1anditislessorequalthantheheightofA.244A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 tion5.A2sistransformable()aPl A:Proof.A2sistransformable()9Ta AandtheexistenceofTa AmeansthatthespreadsofAarepositiveornull,asitistheonlypossiblerestrictiontobuildit.Therefore,a DkP0b DkP09=;()aPl A:Followingthisresult,thetransformationdomainisH Al A;1,whereA2s.SincethelowestlimitofthetransformationisalwayslessorequalthantheheightofA,itisalwayspossibletomakeatransformationforputtinguptheheightofafuzzynumberbut,onthecontrary,thereisaminimumlevelfromwhichtransformationsarenotpossible.InFig.6wehaverepresentedgraphicallythebehaviorofTawhentheheightisdecreasedand,therefore,imprecisionisalsodecreased.Ontheotherhand,inFig.7itisshownhowanincrementofheightproducesanincrementofimprecision.Thisresultagreeswiththefollowingassertion:``Imprecisionanduncertaintycanbeconsideredastwoantagonisticpointsofviewaboutthesamereality,whichishumanimper-fection...andifthecontentsofapropositionismademoreprecise,thenuncer- Fig.6.Transformationthatdecreasesimprecision. Fig.7.Transformationthataugmentsthecertainty.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256245 willhavetobeaugmented''[6],whichisawaytoenunciatetheprincipleofincompatibilitybetweencertaintyandprecision,establishedbyZadehin[17].Consideringthatf AisameasurefortheimprecisionofthefuzzynumberAandthat1ÿh Aisameasureofitsuncertainty,thisprinciplecanbeenunciatedas:·Iff Adecreases,thenh Adecreases.·Ifh Aincreases,thenf Aincreases.FunctionIF®xestheconstantrelationshipbetweenimprecisionandun-certaintyandisassociatedwiththeconceptwerepresentusingafuzzynumber.Ontheotherhand,astransformationstoputuptheheightarealwayspossible,wecanalwaysnormalize(a1)thefuzzynumbersweareworkingwith.Normalizationmeansalossofuncertainty,i.e.,thesecurityonthevalidityofthefuzzyrepresentation.InFig.8itisshownhowthefuzzynumberxkilo-meterswithcertaintydegreelessthan1istransformedintoabiggerfuzzynumberwithcertainty1.Note.Wecanseehow,contrarytothemodelofexpandingimprecisionoverthewholedomain,ourmodelassumesimplicitlythatimprecisionmustbedistributeddependingonthenearnesstotheoriginalconcept.Proposition6(TaProperties).LetA2sanda;b2H A:Thenthefollowingpropertiesareverified:1.Th A AA;2.Ta Tb ATa A;3.Th A Ta AA:Proof.LetusconsiderAf m1;m2;a;b;h Ag,then1.SinceD h A;h A0,thenTh A AA;2.Tb Af m1;m2;aD h A;b;bD h A;b;bg; Fig.8.Anincreaseofcertaintyproducesanincreaseofimprecision.246A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 a Tb Am1;m2;a D h A;bD b;ak;b D h A;bD b;ak;aand,sinceD h A;bD b;aD h A;athenTa Tb ATa A;3.Ittriviallyfollowsfrom(1)and(2).4.2.ThekparameteranditsexperimentalcomputationAswepointedoutattheendofSection3,kparameterisadjusteddependingonthedomainscaletakingintoaccountthatthereisapre-®xedbasescale,forwhichkparameterissetto1.Theideaisthatidenticalfuzzynumbers,thoughexpressedinadierentscale,mustprovideexactlythesameinformationamountandthatthisinformationmustbethesamebeforeandafteratrans-formationisapplied.Inthenextsub-sectionswearegoingtoillustratewithsomeexamplestheuseofkparameterinthecasethatthebasescaleisused(k1)andinthecasewhenitisnot.4.2.1.Fuzzyvaluesinthesamedomainscale:k1LetussupposewearegiventhefuzzynumberAf 3;4;1;1;0:5gfortheconcept`Ibelieveitisfewkilometersfaraway'wherethebelievehasbeenquanti®edby0:5,andwewantAtobenormalized,thatis,a1forthepropositionbecome'Itisfewkilometersfaraway'.Sincetheinformationamountbeforeandafterthetransformationmustbethesame,IF AIF T Aand,byProposition3,D 0:5;1 1ÿ0:5= 10:5.Therefore,thetrans-formationofAisT A3;4;1 1k 1ÿ0:5 10:5;1 1k 1ÿ0:5 10:5;1f 3;4;2;2;1gconsideringk1,sincethetransformationisfromkilometersintokilometers.4.2.2.FuzzyvaluesindierentdomainscaleInthepreviouscase,k1aswewereconsideringa®xedscale,butwhatwouldhappenifweweregiventhesameinformationintwodierentscales?Inthiscase,kissettothenumberofunitsofthebasescalecontainedinaunitofthescaleweareusing.Forexample,ifthebasescaleiskilometersandthescaleinuseismeters,k1=1000,i.e.thenumberofkilometerscontainedinameter.Ifthebasescaleiscentimetersandthescaleinuseismeters,thenk100.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256247 usseetheinformationfunctionbehaviorwithanexample.Inthiscase,wemustestablishabasescaleasareference,forexample,kilometers.Inthissituation,letussupposewewanttonormalizethefollowingfuzzynumbergiveninkilometersandmeters,Akmf 3;4;1;1;0:5ginkilometers;Amf 3000;4000;1000;1000;0:5ginmeters:Akmtransformationgivestheresultobtainedabove:T Akmf 3;4;2;2;1g.ForAmtransformation,kparametermustbesetto1=1000(therelationbetweenthescaleweareusingandthebasescale)andtheresultisT Amf 3000;4000;2000;2000;1gwhichisthesameresultobtainedforT Akmbutrepresentedinthecorrespondingscale.IfweconsidernowthatAkmf 3000;4000;1000;1000;0:5ginkilometers,kparameterwillbe1andthetransformationwillbeT Akmf 3000;4000;1001;1001;1g:Aswecansee,thetransformationfunctionTiscorrectwithrespecttothechangetodierentdomainscales.4.2.3.ExperimentalcomputationofthebasescaleInSection4.2.2,wehaveseenhow,thankstokparameter,wecanusedierentdomainscalesforthefuzzynumberswearehandling.Butthereisanotherkeypointwhenusingtransformationsandthefollowingquestionarises.Shouldtheincrease/decreasebethesameandnotdependonthemeaningofthefuzzynumber?or,inotherwords,shouldtheincreasebethesameirrespectiveofthefactthatwearedealingwithagesorwithdistance?Uptohere,wehaveconsideredthattheusercouldchangesuchincrease/decreasethroughthescalefactork.Inthissectionwearegoingtosee,inanexperi-mentalway,howwecanadjustthetransformationmodeltoeachproblemdomain.Letusthink,forexample,thatwhendealingwithages,theuserispreparedtoadmitthatapproximately40yearsisanyageintheinterval38;42(spreadis2)butwhentalkingaboutdistanceinkilometers,approximately7kilometersisanydistanceintheinterval6;8(spread1).Takingthesecommentsintoaccount,itseemstobereasonablethatkpa-rameterisnotonlydependentonthedomainscalebutalsoontheconceptthefuzzynumberisrepresenting.Inthissense,kshouldhavetheformkk0k;wherekistherelationbetweenthescaleweareusingandthebasescale(asintheprevioussection)andk0isthecorrectionfactorthatdependsonthemeaningandthatallowsustodeterminethereferencescale.Letusseenowhowk0couldbeexperimentallycalculatedconsideringthatweareworkinginabasescale,i.e.k1.248A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 ussupposewehavetherealvaluef x;x;0;0;agwitha1(seeFig.9(a)).Wecouldasktheusertowhatpointheispreparedtorelaxxastoacceptitascompletelytrue.Letussupposetheusersaysthathewillacceptitascompletelytrueifweenlargexatbothsideswithaspreadc,asshowninFig.9(b),becomingtheinitialrealvaluethefuzzynumberf x;x;c;c;1g.Ifitisso,c0 Dk0andD 1ÿa1aand,subsequently,valuek0isk0 1ÿacawithc60:Forexample,ifwehavetherealvaluef 5;5;0;0;0:9gandtheuseradmitsitascompletelytrueifwetransformitintof 5;5;0:5;0:5;1g,asshowninFig.10,thenthecomputedvaluefork0isk0 1ÿ0:90:50:9 0:10:45 14:50:22: Fig.9.Relaxationofxaccordingtouser'scredibility. Fig.10.Relaxationofvalue5forbeingtrue.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256249 then,whereaskdependsonlyonthescalewearedoingthetransfor-mationon,k0permitstheusertospecifythechangesmagnitudedependingdirectlyonthedomainoftheproblemwearetackling.Asaconclusionoftheexperimentalcomputationofk,wecansaythat:1.Thetranslationofcertaintyintoimprecisionisvalidfortheconcreteprob-lemwearetackling,asithasbeenelicitatedexperimentallybytheuser.2.Wecanadjustthetransformationtoparticularproblemsanddomains.3.Itisobviousthattheexperimentforobtainingk0shouldberepeatedmanytimesusingdierentandseparatedomainvaluesanddierentcertaintylev-els.Theideaistoobtainanaveragek0,thatis,ifwedonexperimentsandki0isthevalueobtainedinexperimenti,thenk0Xni1ki0 1n:5.ManagementofuncertainfuzzydataAspointedoutintheintroduction,therearetwomainapproachestodealwithuncertainfuzzynumbers.Fromasemanticalpointofviewandforourpurposes,itseemstobemorereasonabletotruncateatlevelathantoextendthesupportsettothewholedomain.Nevertheless,truncatinghastheincon-veniencethatnon-normalizedfuzzynumbersmustbehandled.Tobeabletoworkalwayswithnormalizedfuzzynumbers,weusethetransformationfunctionintroducedinthispapersettinga1.Thisfunctionwillconvertanuncertainimprecisevalue A;atoanA0whichisnormalizedandprovidethesameinformationastheoriginalAvaluetruncatedtoanalevel.Alltheseprocessescanbesummarizedasfollows:1.WestartfromafuzzynumberAandacertaintylevelaattachedtoit.2.WetruncateAatlevelaassumingthatAheightisitscertaintyvalue.WebuildAa.3.Totakeadvantageofnormalizedfuzzynumberproperties,wetransformAaintoAT(normalizedversionofAausingourtransformationfunction).Wearegoingtoillustratethisprocesswithanexample.LetAbeatrapezoidalnumberexpressedasAf m1;m2;a;b;1gwithanassociatedcertaintylevela.AtruncatedtolevelaisAam1 fÿa 1ÿa;m2b 1ÿa;aa;ba;agandthetransformationofAaisATm1ÿa 1ÿa;m2b 1ÿa;aa 1ÿaka;ba 1ÿaka;1:250A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 canbeprovedthatthedirectexpressionforcomputingATfromAandaisAT 1ÿaa Sa A 1ÿaa Supp A;whereand arethefuzzyextensionsofsumandproductoperators,SthefuzzynumberexpressedbySf 0;0;1=k;1=k;1gandSupp AthesupportsetofAexpressedbythefuzzynumberSupp Af m1ÿa;m2b;0;0;1g.Theadvantageofthisprocessisobvious.Uncertaintyandimprecisionareincludedinthefuzzynumberitselfandthereisnoneedtodeveloporusedierentmechanismsfromthosealreadyintroducedfornormalizedfuzzynumbers.5.1.AnapproximatereasoninginterpretationAswementionedintheintroduction,theproblemoftheco-existenceofbothuncertaintyandimprecisioncanbeformulatedbymeansofthecompo-sitionalruleofinference.Infact,someapproachesthatsolvethisproblemmakeuseofwell-knownimplicationfunctionsonthecertaintydegreeandtheruleconsequent Thestartingpointofourapproachisquitedierentsincewehavetrans-formedAinsuchawaythattheinformationprovidedbyAtruncatedatthecertaintylevelispreserved.Anyway,wearegoingtoprovethatthisapproachisveryclosetotheapproximatereasoningone.Infact,weareprovingthatforthecaseoftrapezoidalfuzzynumbers,thewholeprocessoftransformationappliedisanimplicationfunctioninthesenseofTrillasandValverde[15].Todothat,wearegoingto®ndtheexpressionwhichsummarizesthewholeprocessoftransformationofAwithcertaintyaintoAT.Thisexpressioniscalculatedinthefollowingproperty.Property1.LetAbeatrapezoidalfuzzynumberexpressedasA m1;m2;a;bandletussupposethattheuncertaintylevelofAisa.LetAtruncatedtolevelbeAaandthenormalizedversionofAabeAT.Intheseconditions,itisverifiedthatA.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256251 v2RjlA v0;lAT v1a6lA v;min aaa DklA v DaaD Dk; bba DklA v DbaD Dk();8:whereDD a;1 1ÿa=aandkisthescalefactor.Proof.Provingthispropertyisquitedirecttakingintoaccountthatthea-cutofAhasthefollowingmembershipfunctionlAa vaiflA vPa;lA votherwise:Sothen,ATf m1a;m2a;aa D=k;ba D=k;1g,wherem1am1ÿa: 1ÿam1ÿaaaandm2am2b 1ÿam2bÿba.ThegraphicalrepresentationofthesetrapezoidalfuzzynumbersisshowninFig.11.FromFig.11wecandirectlyobtainthatlAT v1ifa6lA v.Letusseenowthecasewherethisconditionisnotheld.Inthissituation,alA v0andv2 m1ÿa;m1aorv2 m2a;m2ab.Besides, Fig.11.GraphicalrepresentationofA,AaandAT.252A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 A v vÿ m1ÿaaifv2m1ÿa;m1a ; m2bÿvbifv2m2a;m2b :8Finally,lAT v vÿ m1ÿ aÿD=kaaD=kifv2m1ÿ aÿD=k;m1a ; m2bD=kÿvb:aD=kifv2m2a;m2bD=k :8:Makingoperations,wehavelAT v vÿ m1ÿaaa aaD=k D=kaaD=k aaaD=klA v D=kaaD=kifv2m1ÿ aÿD=k;m1a ; m2bÿvbb baD=k D=kbaD=k bbaD=klA v D=kbaD=kifv2m2a;m2bD=k :8Whenv2 m1ÿ aÿD=k;m1athen, aaaD=klA v D=kaaD=k6 bbaD=klA v D=kbaD=kandwhenv2 m2a;m2bD=kthen, bbaD=klA v D=kbaD=k6 aaaD=klA v D=kaaD=k:MakingthesuitableoperationsweobtaintheexpressionofthegiventransformationfunctionlAT v1ifa6lA v;min aaklA vÿa1aa2kÿa1; baklA vÿa1ba2kÿa1:8:Thisresultledustode®neastransformationfunctionforthefuzzynumberAthefunction8x;y2 0;1;I x;y1ifx6y;min axyÿx1ax2ÿx1; bxyÿx1bx2ÿx1:8:Therelationshipofthisresultwiththeapproximatereasoningisshowninthefollowingproperty.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256253 y2.ThefunctionI x;ypreviouslydefinedisanimplicationfunctioninthesenseofTrillasandValverde[15].Proof.1.Iisadecreasingfunctioninx.Thatis8x;x0jx6x0;I x0;y6I x;y.Infact,±ifx;x06ytheresultisimmediate.±ifx6yandx0yitisalsoimmediatesinceI x0;y61.±ifx;x0ytheexpressionis ayxÿx1= ax2ÿx1giventhatxy.Inthiscase,thenumeratorincreaseswithxmoreslowlythanthedenominatorand,therefore,thequotientisadecreasingfunction.2.Iisanincreasingfunctioniny.Thatis,8y;y0jy6y0;I x;y6I x;y0.Thisresultisobvioussincebothfunctions(numeratoranddenominator)arelin-earwithrespecttoywithpositivecoecients.Therefore,thefunctionwillbeincreasinginy.3.I 0;y18y;06yandI 0;y1:4.I 1;yy:Consideringthefollowingexpressionforx1,theresultisdi-rectlyobtained.I x;ymax0;min ayÿ11aÿ11; byÿ11bÿ11y:5.I x;I y;zI y;I x;z(propertyofinterchangeability).Therearetwocases:·y6z.InthiscasetherightsideofthepropertybecomesI x;11byap-plyingproperties1and4,sinceI x;zPI 1;zz.Consequently,y6z6I x;z.Ontheotherhand,theleftsideisalsoequalto1,sincewehavethat:y6z,I y;z1andI x;11,whichprovestheequality.·zy.Therearetwopossibilities:±xPyz.Wearegoingtoprovethat:x6I y;zandy6I x;z,andsubsequentlythatI x;I y;zI y;I x;z1.Infact,thefollowingex-pressionsareheld:xI 1;x6I y;x6I y;zandyI 1;y6I x;y6I x;z.±yzPx.AccordingtoIproperties,I x;z1)I y;I x;z1.WearegoingtoprovethatI x;I y;z1.Todothat,wemustprovethatx6I y;z.Letussupposethecontrary:xI y;z,thenI 1;xI y;zPI y;xandthisleadustotheexpressionI 1;xI y;x;y61whichisincontradictionwiththedecreasingcharacterofI:Sothen,Iisanimplicationfunction.6.ConclusionsTheproblemofimprecisionanduncertaintymanagementthroughfuzzynumbershasbeenaddressed.Fuzzynumbersareausefultoolforrepresenting254A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256 informationbut,inmanycases,thisimpreciseinformationcanbegivenwithanuncertaintydegreeInthesecases,thefuzzynumberincludesadditionalinformationaboutthecon®denceofthisinformation.Inthispaperwehaveproposedamethodthatallowsustotransformthewholeinformation(imprecisionuncertainty)intoanewfuzzynumberinsuchawaythat:·Itmaintainsaprincipleofdistributionoftheimprecisionbasedonametricthattakesintoaccountthedistancetotheoriginalconcept,thatis,thecloseranelementistotheconceptthemoreistheincreaseofitsmembershiptothementionedconcept(basedonZadeh'sprinciple).·Itpermitstoadjusttheresultstotheusers'pointofviewbyusingthescalefactor,i.e.wecanadjustthetransformationtoparticularproblemsanddo-mains.·Itisinterpretedfromtheapproximatereasoningpointofviewand,subse-quently,itguaranteessoundresults.·Itiseasytoimplementand,therefore,tobeincludedasadatapreprocessingmodule.·Themethod,notonlynormalizesbutequalizes.Thisisanimportantchar-acteristicforproblemswhereanacceptedlevelofuncertaintyexists(notnec-essarily1).·Ifbissetto1,allthesoftwaredevelopedfornormalizedfuzzysetsisre-us-ableandnonewversionsarenecessarytotreatuncertainty.Asfutureavenuesforresearchwecanmention:1.Togiveageneralexpressionforfuzzynumbers.Inthispaperwehaveonlyconsideredthecaseoftrapezoidalfuzzynumbersbuttheresultscouldbegeneralizedforanykindoffuzzynumber.2.Touselinguisticuncertaintyinsteadofuncertaintylevels.3.Tostudyhowtransformationsaecttheresultsobtainedfromarithmeticoperations,matchingorrankingoffuzzyvalues.References[1]G.Bordogna,G.Pasi,Managementoflinguisticquali®cationofuncertaintyinafuzzydatabase,in:ProceedingsofIPMU'98,SeventhInternationalConferenceonInformationProcessingandManagementofUncertaintyinKnowledge-basedSystems,Paris,France,1998,pp.435±442.[2]Z.Cao,D.Park,A.Kandel,Investigationsontheapplicabilityoffuzzyinference,FuzzySetsandSystems49(1992)151±169.[3]J.L.Castro,J.J.Castro,J.M.Zurita,Thein¯uenceofimplicationfunctionsinfuzzylogiccontrollerswithdefuzzi®cation®rstandaggregationafter,in:SeventhIFSAWorldCongress,Prague,1997,pp.237±242.[4]M.Delgado,J.L.Verdegay,M.A.Vila,Onvaluationandoptimizationproblemsinfuzzygraphs:Ageneralapproachandsomeparticularcases,ORSAJournalonComputing2(1990)74±83.A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256255 M.Delgado,M.A.Vila,W.Voxman,Onacanonicalrepresentationoffuzzynumbers,FuzzySetsandSystems93(1998)125±135.[6]D.Dubois,Modelesmathematiquesdel'imprecisetdel'incertainenvued'applicationsauxtechniquesd'aidealadecision,Ph.Dthesis,UniversiteScienti®queetMedicaledeGrenoble,1983.[7]D.Dubois,H.Prade,Fuzzynumbers.anoverview,in:Bezdek(Ed.),TheAnalysisofFuzzyInformation,CRSPress,BocaRaton,1985.[8]D.Dubois,H.Prade,PossibilityTheory,PlenumPress,NewYork,1985.[9]D.Dubois,H.Prade,Themeanvalueofafuzzynumber,FuzzySetsandSystems24(1987)279±300.[10]J.A.Goguen,L-fuzzysets,JournalofMathematicalAnalalysisandApplications18(1967)145±174.[11]A.Gonzalez,MetodosSubjetivosparalaComparaciondeNumerosDifusos,Ph.Dthesis,GranadaUniversity,1987.[12]A.Gonzalez,Astudyoftherankingfunctionapproachthroughmeanvalues,FuzzySetsandSystems35(1990)29±41.[13]KampedeFeriet,J.Latheoriegeneraliseedel'informationetlemesuresubjetivedel'information,in:KampedeFeriet,C.F.Picard(Eds.),LecturenotesinMath:Theoriesdel'information,vol.398,Springer,Berlin,1973,pp.1±28.[14]O.Pons,J.C.Cubero,A.Gonzalez,M.A.Vila,Uncertainfuzzyvaluesstillintheframeworkof®rstorderlogic,in:ProceedingsofSecondInternationalWorkshoponLogicProgrammingandSoftComputing:TheoryandApplications,Manchester,UK,1989.[15]E.Trillas,L.Valverde,Onimplicationandindistinguishabilityinthesettingoffuzzylogic,J.Kacpryzk,R.R.Yager(Eds.),ManagementDecisionSupportSytemsusingFuzzySetsandPossibilityTheory,VerlagTUVRheinlan,1985,pp.189±212.[16]R.R.Yager,Measurementofpropertiesonfuzzysetsandpossibilitydistributions,E.P.Klement(Ed.),ProceedingsoftheThirdInternationalSeminaronFuzzySetTheory,JohannesKeplasUniversityLiuz,1981,pp.211±222.[17]L.A.Zadeh,Outlineofanewapproachtotheanalysisofcomplexsystemsanddecisionprocesses,IEEETransactionsSystems,ManandCybernetics3(1973)28±44.256A.Gonzalezetal./Internat.J.Approx.Reason.21(1999)233±256