Applying consistent fuzzy preference relations to partnership selection.docx
《Applying consistent fuzzy preference relations to partnership selection.docx》由会员分享,可在线阅读,更多相关《Applying consistent fuzzy preference relations to partnership selection.docx(17页珍藏版)》请在冰豆网上搜索。
Applyingconsistentfuzzypreferencerelationstopartnershipselection
Applyingconsistentfuzzypreferencerelationstopartnershipselection
Tien-ChinWanga,1,
andYueh-HsiangChenb,c,
aDepartmentofInformationManagement,I-ShouUniversity,Kaohsiung840,Taiwan
bDepartmentofInformationEngineering,I-ShouUniversity,Kaohsiung840,Taiwan
cDepartmentofInformationManagement,Kao-YuanUniversityofTechnology,Kaohsiung821,Taiwan
Received28January2005;
accepted19July2005.
Availableonline9September2005.
Abstract
Partnershipselectionhasbeenimportanttotheformationofavirtualenterprise.Basedonthefuzzypreferenceprogramming(FPP)methodproposedbyMikhailov[Fuzzyanalyticalapproachtopartnershipselectioninformationofvirtualenterprises.Omega2002;30:
393–401],thisinvestigationpresentsaconsistentfuzzypreferencerelationsmethodtoselectpartners.Humanthoughtsarefullofuncertainty,sothedecision-makerscannotmakeexactpairwisecomparisons.TheFPPmethodsolvesthisproblemusinganintervalvalueinsteadofSaaty's1–9scale.Inthisstudy,theFPPmethodisreviewed,andthentheconsistentfuzzypreferencerelationsmethodiselucidated.Finally,thepresentedmethodisappliedtotheexampleaddressedbyMikhailov[Fuzzyanalyticalapproachtopartnershipselectioninformationofvirtualenterprises.Omega2002;30:
393–401].Thisstudyrevealsthattheproposedmethodyieldsconsistentdecisionrankingsfromonlyn-1pairwisecomparisons—thesamenumberasinMikhailov'sresearch.Briefly,thepresentedconsistentfuzzypreferencerelationsmethodisaneasyandpracticalwaytoproviderankingsofpartnershipinmakingdecision.
Keywords:
AHP;FPP;Consistency;Consistentfuzzypreferencerelations
ArticleOutline
1.
Introduction
2.
FPPmethod
3.
Consistentfuzzypreferencerelations
4.
Illustrativeexample
5.
Conclusions
References
1.Introduction
AccordingtoMikhailov's[1]researchonpartnershipselection,allrelationalissuescanbeaddressedbytraditionalAHPmethod.However,thetraditionalAHPmethodisproblematicinthatitusesanexactvaluetoexpressthedecisionmaker'sopinioninacomparisonofalternatives.However,inreality,ahumanpreferencemodelisuncertain.Inthepartnershipselectionprocess,thedecisionmakerisgenerallyunsureofhispreferencesbecauseinformationaboutthefuturepartnersandtheirperformanceisincompleteanduncertain.Someofthedecisioncriteriaaresubjectiveandqualitative,sothedecisionmakercannoteasilyexpressthestrengthofhispreferencesorprovideexactpairwisecomparisons.MikhailovutilizestheintervalvaluestoexpressthecomparisonsanddevelopstheFuzzyPreferenceProgramming(FPP)methodtocalculatetheweightofeverylevel,whichiscalledalocalpriority,tosolvetheproblem.HethenusedtheAHPmethodtodeterminetheglobalprioritiesbyaggregatingallofthelocalpriorities.Finally,heappliedthemethodinanexampletorankthealternativesandselections.
TheFPPmethodprovidessomeadvantages,includingaconsistencyindicator,simplicityofcomputation,highprecisionandpreservationofranks.BasedontheFPPmethod,thispaperappliesthefuzzypreferencerelationsmethodproposedbyHerrera-Viedmaetal.[2]toselectthevirtualpartners.Themethodconstructsthedecisionmatricesofpairwisecomparisonsusinganadditivetransitivity.Onlyn-1comparisonsarerequiredtoensureconsistencyforalevelwithncriteria.Themethodissimplyandpracticallyprovidesrankingchoicesindecision-makingproblems.
Thispaperisorganizedasfollows.Section2reviewstheFPPmethod.Section3reviewstheconsistentfuzzypreferencerelations.Section4presentsanillustrativeexample,andSection5drawstheconclusions.
2.FPPmethod
MikhailovproposedtheFPPmethod[1],[2],[3]and[4]toderivepriorityvectorsfromasetofcrisporintervalcomparisons.Theassessmentoftheprioritiesisanoptimizationproblem,maximizingthedecision-maker'ssatisfactionwithaspecificcrisppriorityvector.
ConsidertheintervalpairwisecomparisonswithncriteriaatthesamelevelinahierarchyX1,X2,…,Xn.aij=(lij,uij)isassumedtorepresentthedecision-maker'sintervaljudgmentsofcriteriaXitoXj.Whentheintervaljudgmentsareconsistent,severalpriorityvectors,whosecomponentratiossatisfytheinequalitieslij
wi/wj
uij,i=1,2,…,n-1,j=2,3,…,n,j>i.Nopriorityvectorsimultaneouslysatisfiesallintervaljudgmentsintheinconsistentcase.However,avectorthatsatisfiesalljudgments“aswellaspossible”canbereasonablysought,indicatingthatasufficientlygoodsolutionvectormustapproximatelybeconsistentwithalljudgments,or
(1)
where
denotesthestatement“fuzzylessorequalto”.
Theinequality
(1)canbetransformedintoasetoftwosingle-sidefuzzyconstraintstohandleeasily,
(2)
Theabovesetofn×(n-1)fuzzyconstraintscanbepresentedinamatrixformas,
(3)
wherethematrix
m=n×(n-1).
Thekthrowof
(2),forwhich
k=1,2,…,m,representsafuzzylinearconstraintandisdefinedbyalinearmembershipfunctionofthetype
(4)
wheredkisatoleranceparameter,whichrepresentstheadmissibleintervalofapproximatesatisfactionofthecrispinequalityRkw
0.Themembershipfunction(4)representsthedecision-maker'ssatisfactionwithaspecificpriorityvector,withrespecttothekthsingle-sideconstraint
(2).Thevalueofthemembershipfunctionμk(Rkw)iszerowhenthecorrespondingcrispconstraintRkw
0isstronglyviolated;itriseslinearlyandtakespositivevaluesoflessthanonewhentheconstraintisapproximatelysatisfied.Ittakesvaluesofgreaterthanonewhentheconstraintisfullysatisfied.
Letμk(Rkw),k=1,2,…,mbethemembershipfunctionsofthefuzzyconstraints
onthe(n-1)dimensionalsimplex
.
Definition 2.1
Thefuzzyfeasiblearea
onthesimplexTn-1isafuzzyset,
(5)
Thefuzzyfeasibleareaisdefinedasanintersectionofallfuzzyconstraintsonthesimplex.Ifthetoleranceparametersoftheirfuzzysetsare“largeenough”,thenanon-emptyfuzzyfeasibleareacanbeobtained.Therefore,anon-emptyfeasiblearea
onthesimplexTn-1isaconvexfuzzyset.
Theconvexfuzzyfeasiblearea
representstheoverallsatisfactionofthedecision-makerwithaspecificcrisppriorityvector.Assumingthatthedecision-makerisinterestedinthebestpossiblesolution,apriorityvectorthatmaximizeshisoveralldegreeofsatisfactioncanbereasonablydetermined.
Definition 2.2
Themaximizingsolutionisavectorw*,whichcorrespondstothemaximumfuzzyfeasiblearea
(6)
Thefuzzyfeasiblearea
isaconvexsetandallfuzzyconstraintsaredefinedasconvexsets,soatleastonepointw*isalwayspresentonthesimplexthathasamaximumdegreeofmembershipin
.
Theproblemoffindingthemaximizingsolutionistransformedtoalinearprogrambyintroducingavariableλ;measuringthedegreeofmembershipofthefuzzyfeasiblearea
andusingEqs.(4)and(6);
(7)
TheoptimalsolutiontoEq.(7)isavector(w*,λ*)whosefirstelementrepresentsthepriorityvectorthathasamaximumdegreeofmembershipinthefuzzyfeasiblearea,whilethesecondelementrepresentsthevalueofthatmaximumdegree,
.Thevalueofλ*representsthedegreeofsatisfactionandisanaturalindicatoroftheinconsistencyofthedecision-maker'sjudgments,andsocanberegardedasaconsistencyindex.Whentheintervaljudgmentsareconsistent,λ*isone.Forinconsistentjudgments,λ*isbetweenoneandzero,dependingonthedegreeoftheinconsistencyandthetoleranceparametersdk.
3.Consistentfuzzypreferencerelations
Herrera-Viedmaetal.proposedtheconsistentfuzzypreferencerelations[2]forconstructingthedecisionmatricesofpairwisecomparisonsbasedonadditivetransitivity.Fuzzypreferencerelationsenableadecision-makertogivevaluesforasetofcriteriaandasetofalternatives.Thevaluerepresentsthedegreeofthepreferenceforthefirstalternativeoverthesecondalternative.Twomajorkindsofpreferencerelationsapply—
(1)multiplicativepreferencerelations,and
(2)fuzzypreferencerelations.
(1)Multiplicativepreferencerelations[5]:
AmultiplicativepreferencerelationRintermsofasetofalternativesAisrepresentedbyamatrixR
A×A,R=(rij),whererijisthepreferenceratioofalternativeaitoaj.Saatysuggestsmeasuringrijusingaratioscale,andthedefined1–9scale.Herein,rij=1representstheabsenceofadifferencebetweenaiandaj;rij=9denotesthataiismaximallybetterthanaj.Inthiscase,thepreferencerelationRistypicallyassumedtobeamultiplicativereciprocal,
j
{1,…,n}.
(2)Fuzzypreferencerelations[6],[7]and[8]:
AfuzzypreferencerelationPonasetofalternativesAisafuzzysetontheproductsetA×Awithmembershipfunctionμp:
A×A→[0,1].Thepreferencerelationisrepresentedbythen×nmatrixP=(pij),where
.Herein,pijisthepreferenceratioofalternativeaito
meansthatnodifferenceexitsbetweenaiandaj,pij=1indicatesthataiisabsolutelybetterthanaj,and
indicatesthataiisbetterthanaj.Inthiscase,thepreferencematrixPisgenerallyassumedtobeanadditivereciprocal,pij+pji=1,
i,j
{1,…,n}.
Herrera-Viedmaetal.[2]proposedaconsistentadditivepreferencerelation.Thesepropositionsaredescribedasfollows.
Proposition 3.1
Considerasetofalternatives,X={x1,…,xn},associatedwithareciprocalmultiplicativepreferencerelationA=(aij)foraij
[1/9,9].Then,thecorrespondingreciprocalfuzzypreferencerelation,P=(pij)withpij
[0,1]associatedwithAisgivenas
.
isconsideredbecauseaijisbetween
an