Thegeneralizationnegationof probabilitydistributionandits applicationintargetrecognitionbased onsensorfusion
XiaozhuanGaoandYongDengAbstract
InternationalJournalofDistributed SensorNetworks 2019,Vol.15(5)
TheAuthor(s)2019 DOI:10.1177/1550147719849381 journals.sagepub.com/home/dsn
Targetrecognitioninuncertainenvironmentsisahotissue.Fusionrulesareusedtocombinethesensorreportsfromdifferentsources.Inthissituation,obtainingmoreinformationto makecorrectdecisionisanessentialissue.Probabilitydistributionisoneofthemostusedmethodstorepresentuncertaintyinformation.Inaddition,thenegationofprobability distributionprovidesanewviewtorepresenttheuncertaintyinformation.Inthisarticle,theexistingnegationofprobability distributionisextendedwithTsallisentropy.Themainreasonisthatdifferentsystemshavedifferentparameter q.Some numericalexamplesareusedtodemonstrate theefficiencyoftheproposedmethod.Besides,thearticlealsodiscussesthe applicationofnegationintargetrecognitionbasedonsensorfusiontofurtherdemonstratetheimportanceofnegation.
Keywords
Probabilitydistribution,negation,Tsallisentropy,Ginientropy,sensorfusion
Datereceived:5January2019;accepted:17April2019
HandlingEditor:MohamedAbdel-Basset
Introduction
Inrecentyears,informationfusionispaidgreatattentioninmilitaryapplications.1,2 Manymethodsbased oninformationfusionhavebeenproposedtoclassify objects,3 targetrecognition4,5 anddecisionsmaking.6,7 Inmostcases,thefinalresultwhichisobtainedby informationfusionofmultiplesensorsmaybereasonable.However,therecognitionresultmaybecounterintuitiveduetohighconflict.8 Uptonow,fusionruleis stillanopenquestion.Besides,theinformationgatheredinsensorsfusingsystem9–11 existsuncertainasitis incomplete,inconsistentandpossiblyimprecise.Many methodshavebeenproposedtoobtainmoreinformation.12–14 However,duetotheexistinguncertainty,15–17 itisessentialtoobtainmoreinformationaccordingto knownknowledge.
Presentingknowledgeisanopenissue.18–20 Inmost cases,wecommonlyuse‘must’,‘may’and‘likely’to
estimatewhetheraneventwillhappenornot.Dueto theuncertainty,therearesomemethodstodealwith it.21–24 Probabilitydistributionisusedtoquantitatively describethepossibilityofoccurrenceofaresultinreal applications.25 Moreimportantly,formostcases,itis mucheasiertodescribethenegationoftheeventsthan directlydescribetheminsomecircumstances.For example,ifitisdifficulttoproveamathematicalformularigorously,however,acounterexamplecaneasily provetheformulawrong.Similarly,thereisasignificantpropertyinprobability,namelymutualexclusion.
InstituteofFundamentalandFrontierScience,UniversityofElectronic ScienceandTechnologyofChina,Chengdu,China
Correspondingauthor: YongDeng,InstituteofFundamentalandFrontierScience,Universityof ElectronicScienceandTechnologyofChina,Chengdu610054,China. Emails:dengentropy@uestc.edu.cn;prof.deng@hotmail.com
CreativeCommonsCCBY:ThisarticleisdistributedunderthetermsoftheCreativeCommonsAttribution4.0License (http://www.creativecommons.org/licenses/by/4.0/)whichpermitsanyuse,reproductionanddistributionoftheworkwithout furtherpermissionprovidedtheoriginalworkisattributedasspecifiedontheSAGEandOpenAccesspages(https://us.sagepub.com/en-us/nam/ open-access-at-sage).
Mutualexclusionmeanstheoccurrenceofonething whenitsoppositewillnothappen.Inotherwords,the probabilityofaneventwillaffectitsopposition. Therefore,itismeaningfultostudynegationofprobabilitydistribution(NPD).26 Recently,theNPDbased onGinientropywasproposedbyYagertopresent knowledgeinanewview.27
Themotivationofthisstudyistofindamoregeneralandreasonablemodetoevaluatetheuncertainty ofNPD.Ifthecorrelationsbetweenthe N elementsare strongenough,thentheextensivityofsomeentropiesis lost,whichisincompatiblewithclassicalthermodynamics.28 Tsallisentropy29 isproposedtoovercome thisdifficultywithnon-extensiveproperty.InTsallis entropy, q isnota‘tunable’parameter,butisstrictly determinedbythesystemitself.30 Asaresult,always usingGinientropyinYager’sNPD,27 whichisaspecial caseofTsallisentropywhen q = 2,isnotreasonableto manyrealsystems.Tosolvethisproblem,thearticle usesTsallisentropytomeasuretheuncertaintyof NPD,whichcanbemorereasonable.
Moreimportantly,innatureandsociety,everything hasitsnegation.Regretandexpectgiveustwoviewsto consideraproblem.Moreover,thebestalternativeisas closeasidealsolutionandisasfarasnegativesolution. Besides,NPDcanprovideanotherinformationbased onknowninformation.Thatistosay,wecananalyse thistargetfromtwosides,whichcanimprovethecorrectnessofdecision-making.ThereasonisthatNPD hasthepropertiesofimprecisionandunknown,which isbeneficialfordecision.Moreimportantly,iftheoriginalinformationishighlyconflicting,theconflictof negationmaynotbehighlyconflicting.Basedonthe discussion,itisofgreatsignificancetostudynegation. Hence,thearticlealsousesnegationtorecognizetarget basedonsensorfusion.
Therestofthisarticleisstructuredasfollows:Inthe section‘Preliminaries’,preliminariesofsomeentropies andYager’snegationmethodareintroduced.Theproposedmethodisintroducedinthesection‘Theproposedmethod’.Inthesection‘Examplesand discussion’,anumericalexampleisusedtoillustrate themethodofnegationandusingTsallisentropyto evaluatetheuncertainty.Theapplicationofnegationin targetrecognitionbasedonsensorfusionisintroduced inthesection‘Applicationofnegationbasedonsensor fusion’.Finally,someconclusionsaregiveninthesection‘Conclusion’.
Preliminaries
Inthissection,thepreliminariesofsomeentropiesand NPDwillbebrieflyintroduced.
Ginientropy
Entropyplaysaveryimportantroleinmanysystems.31–33 InYager’sNPD,theGinientropyisadopted.27
Definition1. Ginientropyisdefinedasfollows34
HG = X n i = 1 P(Ai )(1 P(Ai )) ð1Þ
where pi istheprobabilitydistributionwhichsatisfies Pn i = 1 Pi = 1.Ginientropyisexpansive,thatis, G (P1 , P2 , , Pn )= G (P1 , P2 , , Pn , 0).Itmeansaddinganelementwithzeroprobabilitydoesnotaffectthe Ginientropy.
Moreover,ifonlyforthesakeofcomparingtheuncertaintyassociatedwiththetwodistributions,theGini entropymaybepreferredtotheShannonwithsimplercalculation.35 Knowledgerepresentationisofgreatsignificance tomodernscience.Inmanyfields,ithasbeenregardedas themaindrivingforcefortheapplicationoftheorytopractice,suchas,aggregation,36 evidenceresolution,37–39 decision-making40–42 andsoon.43–45 Interestingly,probabilityis similartocoins,italsohasaoriginalsideandanegative side.46 Summarizing,negationprovidesanewviewtoinvestigatethepropertyofprobability.
Definition2. Assumingaprobabilitydistribution P = P(A1 ), P(A2 ), , P(An ) fg,theNPDwasdefined as 27 P(Ai )= 1 Ai n 1 ð2Þ
Becausetheprobabilitiesarecompletelymutually exclusive,itsnegationwasnormalized,sotheNPD satisfies 0 ł P(A1 ) ł 1 X P(Ai )= 1
Itshouldbepointedthatsincethebasicprobability assignmenthasmoreflexibilitytorepresentuncertainty, thenegationofbasicprobabilityassignmentisalsopaid attentionrecently.46
Tsallisentropy
Definition3. Givenaprobabilitydistribution P = P1 , P2 , Pn fg,theTsallisentropycanbedefined as 29
Hq P ðÞ = k q 1 1 X i Pq i !
when q ! 1,whichcorrespondstotheBoltzmannGibbsentropy,namely
H (P)= H1 (P)= X i 1 Pi lnPi ð4Þ
OnecanrewriteTsallisentropyasfollows29
Hq = k q 1 X W 1 Pi (1 Pq 1 i ) ð5Þ
where q isarealparametersometimescalledentropicindexwhichgeneralizestheusualexponentialandplays theimportantroleindescriptionofthermodynamic. k istheBoltzmannconstant(orsomeotherconvenient value,inareasoutsidephysics,suchasinformationtheory,cyberneticsandothers).28,29 Besides,when q = 2, theTsallisentropydegeneratestheGinientropy.
Theproposedmethod
Fromtheabove,itcanbeseenthatYager’smethod wouldbethespecialcase.Inordertoexpandtheapplicationofnegation,thearticleproposedamethodthat usesTsallisentropytomeasuretheuncertaintyof NPD.UsingTsallisentropytomeasuretheuncertainty ofNPDcanexpandtheapplicationofnegation
Hq = 1 q 1 X 1 Pi n 1 1 ( 1 Pi n 1 ) q 1 ð6Þ
Inthefollowingpart,thepropertyofnegation-based proposedmethodcanbediscussed.
Property1. Tsallisentropycanincreaseafternegation. Specificproofisasfollows Hq (P)= 1 q 1 3 (1 X (Pq i )) Hq (P)= 1 q 1 3 (1 X (Pq i )) when q ¼ 1 Hq (P) Hq (P)= 1 q 1 3 ( X Pq i X 1 Pi n 1 q ) D0 = Hq (P) Hq (P)+ l X N i = 1 Pi 1! ∂D0 ∂Pi = q q 1 pq 1 + q q 1 1 n 1 q 1 Pi ðÞq 1 + l Hq (P) Hq (P) ø 0 when q = 1
H1 (P) H1 (P)= X (1 Pi )ln(1 Pi )+ X Pi lnPi
D0 = H1 (P) H1 (P)+ l X N i 1 Pi 1!
∂H1 (P) H1 (P) ∂Pi = lnPi + 1 n 1 3 ln 1 Pi n 1 + l H (P) ø H (P)
Fromtheabovecalculation,itcanbeseenthat Tsallisentropycanincreaseafternegation.
Property2. ThereismaximumTsallisentropywithuniformdistributionaftermultiplenegations.
Thegeneralformulaofthenegationmethodisas follows xi + 1 = 1 n + n 3 x1 1 n 3 (n 1)i 1 ð7Þ
where xi + 1 representsthe ithnegation.Aftermultiple negations,theprobabilitydistributionisasfollows lim i!‘ xi + 1 = 1 n
Fromtheabove,itcanbeseenthatTsallisentropy reachesthemaximumaftermultiplenegations.
Besides,innature,anyoperationcancauseenergy consumption.Thechangeofentropymeanstheconsumptionofenergy,andconsumedenergycannotbe reused.Similarly,thevariationofinformationentropy isaccompaniedbytheconsumptionofinformation.Of variousentropies,Tsallisentropyisnon-extensive entropyandappliedinartificialandsocialcomplexsystems.Hence,usingTsallisentropytomeasureuncertaintyofNPDcanexpandtheapplicationsofnegation.
Examplesanddiscussion
Example
Comparing
whichattachestolowerprobabilityacquireshigherprobabilityafternegationprocess. Attractively,whathappensifnegationistakento P(xi ) ?Secondnegationismadeandtheresultsareobtained
Itisapparentthattheoriginalprobabilitydistribution P(xi )isnotequalto P(xi ).Generally,thisprocess forobtainingtheNPDisirreversiblewhen N ø 3
P(xi ) ¼ P(xi )
Itisnecessarytoexplorewhatcausesthisirreversibility.FromExample1,itcanbefoundthattheprobabilitywouldberedistributionafternegation.Hence,it shouldbeconsideredwhethertheuncertaintyhaschangedaftertakingNPD.Inthenextsection,thereare somediscussionsonwhythenegationprocessisgenerallyirreversibleandhowtomeasuretheuncertaintyof theprobabilitybasedonournegationmethod.
Furtherdiscussion
AgainconsideringtheExample1,theTable1isused toshowthechangeofprobabilityaftereachiteration ofnegationprocess.Itcanbeclearlyknownthatprobabilityisreallocatedafternegation.Gradually,the probabilitydistributionbecomesmoreandmoreclose totheuniformdistribution.Whatisthecauseofthis phenomenon?
Theconceptofentropyisderivedfromphysics.47,48 Withtheincreasingapplicationofentropy,49,50 informationentropyhasbecomeanindispensablepartof modernscientificdevelopment.51,52 Shannon53 first proposedtheconceptofinformationentropyto describetheuncertaintyandhasappliedinalotsof fields.54,55 However,asintypicalphysicalproblems, therearesomeexampleswheretheBoltzmannShannonentropyisnotsuitable.56,57 In1988,Tsallis proposedanon-extensiveentropycalledTsallis entropy.Subsequently,non-extensivestatistical mechanicswhichisGeneralizationofBoltzmannGibbsStatisticsemergedbasedonTsallisentropy. Moreimportantly,theBoltzmann-Gibbsstatisticsis recoveredasthelimitationwhen q ! 1 58 Thevalueof q ishiddeninthemicroscopicdynamicsofthesystem. Itcanbeobtainedbysomeexperiments.Arelevant improvementontheinferenceaccuracybyadopting non-extensiveentropiesisproposedbyTsallis,57 where q equals2.5obtainsabetterresult.Besides,in article,59q = 2 5 isusedtoobtainmutualinformation duringDREAM4data.Asaconsequence,usingTsallis entropytomeasuretheuncertaintyextendedthe methodofYager.27 Therefore,itisofgreatsignificance tomeasureuncertaintyofnegationwithTsallis entropy.
Let’scalculatetheuncertaintyusingTsallisentropy withdifferent q toobservehowTsallisentropychange afternegation,asshowninFigure1.Therearechanges ofTsallisentropywithfivedifferent q from 1 2 to3.It canbeeasilyseenthattheTsallisentropyincreases
Figure1. ThetrendofTsallisentropyaftereachiterationof negationprocess.
Table1. Thechangeofprobabilityaftereachiterationof negationprocess.
Frequencyofnegation P (x1 ) P (x2 ) P (x3 ) 00.20000.50000.3000 10.40000.25000.3500 20.30000.37500.3250 30.35000.31250.3375 40.32500.34370.3313 50.33750.32810.3344 60.33100.33600.3330 70.33450.33200.3335 80.33300.33400.3330 90.33330.33320.3335 100.33330.33330.3334
graduallyintheprocessofnegationandisalmost invariableafterthefifthnegation,regardlessofthe valueof q.Itisclearlyshownthattheuncertaintygraduallyincreasesinanysystemaftereachiterationof negationprocess.Whentherearemultipleiterationsof negationprocess,theprobabilitydistributionwouldbe uniformdistributionandTsallisentropyhasmaximum entropy,whichisconsistentwith Property 1and Property 2.Hence,usingTsallisentropytomeasure uncertaintyofNPDisreasonable.
Applicationofnegationbasedonsensor fusion
Targetrecognitionispaidgreatattentioninmilitary applications.60 Therearesomemethodstomakedecisions.61–63 Fusionrulescanhelpusmakebetterdecisions.64,65 However,usingexistinginformationtomake
moreaccuratedecisionisalsoanopenissue.Thenegationprovidesanewviewtoobtainmoreaccuratejudgementwiththecollectedinformation.Thespecific exampleisasfollows.
Therearethreesensorstorecognizethetargetwhich maybe A, B and C.Theresultsfromthesensorsareas follows
p1 (A)= 0 7, p1 (B)= 0 2, p1 (C )= 0 1 p2 (A)= 0:6, p2 (B)= 0:3, p2 (C )= 0:1 p3 (A)= 0 5, p3 (B)= 0 3, p3 (C )= 0 2
Dempsterruleiswidelyusedinsensordatafusion.64 Thecombinationresultsareshownasfollows
K = 0 77, p(A)= 0 91, p(B)= 0 078, p(C )= 0 087
Usingthemethodofnegation,newinformationis obtainedasfollows
p1 (A)= 0 15, p1 (B)= 0 4, p1 (C )= 0 45 p2 (A)= 0 2, p2 (B)= 0 35, p2 (C )= 0 45 p3 (A)= 0 25, p3 (B)= 0 35, p3 (C )= 0 4
Similarly,fusionruleisusedasfollows
KN = 0 8625, p(A)= 0 055; , p(B)= 0 356, p(C )= 0 375
Itiswellknownthatconflictingcoefficient K plays anessentialroleininformationfusion.8 Asaresult, analysingthedifferenceof K betweentheoriginalprobabilityanditscorrespondingnegationisnecessary.Itis easilyfoundthattheconflictafternegationbecomes bigger.Thereasonisthat p(A)= 1 p(A)represents theprobabilitythetargetisnot A,namely p(A)isthe probabilitythattargetis B or C.Besides,themethod ofnegationcontainstheuncertainclass(AorB).Hence, theconflictbecomesbiggerduetotheexistinguncertainclassafternegation.
Moreimportantly,bycomparingtheresultsbetween theoriginalandnegation,thedecisionhasmoresupporttotarget A p(A)= 0 055 reflectstheprobability thatthetargetisnot A.Thatistosay,itcanincrease theprobabilityoftarget A fromtheotherside.Hence, itcangettworesultsfromthetwosidesbasedondata, whichisbettertomakeamorereasonabledecisionin targetrecognition.
Next,consideringthechangesofentropybetween theoriginalandnegation,assumethat q = 3 asthere arethreesensors
T1 = 0 324, T2 = 0 378, T3 = 0 42, T = 0 123
T 1 = 0:421, T 2 = 0:429, T 3 = 0:439, T = 0:375
Fromtheabove,itcanbeseenthattheTsallis entropyafterfusionbecomesless,showingthatthe fusioncandecreasetheuncertaintyofinformation.
However,fusionrulesdon’tworkinextremeprobabilitydistribution.Usingnegationcanprovideanother viewtoanalysethisphenomenon.
Thereisaspecialexampletobetterexplaintheapplicationofnegation.Therearetwosensorstorecognize thetargetswhichmaybe A, B and C.Theresultsfrom thesensorsareasfollows
p1 (A)= 0 8, p1 (B)= 0, p1 (C )= 0 2 p2 (A)= 0, p2 (B)= 0 8, p3 (C )= 0 2 p2 (A)= 0, p2 (B)= 0:8, p3 (C )= 0:2
Usingfusionrules,onecangettheresultsasfollows
p(A)= 0, p(B)= 0, p(C )= 1
Fromtheresult,itcanbeseenthatthetargetmust be C as p(C )= 1.Obviously,theresultisnotveryreliable.Besides,itcanbeseenthatwhenevidenceishighly conflicting,Dempsterruledoesn’twork.However, negationcanprovideanotherinformation,whichis usefultomakedecision
p1 (A)= 0 1, p1 (B)= 0 5, p1 (C )= 0 4 p2 (A)= 0 5, p2 (B)= 0 1, p2 (C )= 0 4
Thefusionresultsareasfollows
p(A)= 0:19, p(B)= 0:19, p(C )= 0:62
Next,consideringthechangesofentropybetween theoriginalandnegation,assumethat q = 2 asthere aretwosensors
T1 = 0 4800, T2 = 0 4800, T = 0 T 1 = 0 8100, T 2 = 0 8100, T = 0 7480
Obviously,negationgivesussomenewinformation. Fromtheaboveresult,theprobabilitythattargetisnot C is0.62.Itcanbeseenthattheresultofnegationis morereasonablethantheoriginalresult.Moreimportantly,negationprovidesanotherviewtoanalyseproblemandcanbetterhandleconflictingevidence.Hence, negationisusefultomakedecision.
Inaddition,analysingtheTsallisentropybetween originalandnegation,itcanbefoundthatTsallis entropybecomesbigger.Besides,accordingtothesecondlawofthermodynamics,theentropyofanisolated systemneverdecreases.Thenegationcanincrease Tsallisentropy.Moreover,entropyisirreversible. Fromthisview,thenegationcanmakesystemsspontaneouslyevolvetowardsthermodynamicequilibrium,
thestatewithmaximumentropy.Thatistosay,negationtendstomakethesystemuniformlydistributed, andinfactitdoes.Hence,thenegationnotonlyprovidesanewwaytounderstandproblems,butalsoprovidesbetterperformanceofdecisionsupportsystem. Besides,becausetheTsallisentropyishighlyappliedin manyapplications,negationprovidesanewviewto obtaininformation.Hence,usingTsallisentropyto measuretheuncertaintycanenlargetheapplicationof negation.
Conclusion
Probabilitydistributionisefficienttorepresentknowledge.However,everythinginnatureandsocietyhasits negation,whichshowsthatnegationisveryessential. Similarly,probabilitydistributionalsohasitsnegation. Thearticleextendstheproposednegationbyusing Tsallisentropy.Besides,numericalexampleisusedto calculatetheuncertaintybydifferent q ofTsallis entropyafternegation.Itcanbefoundthatthereis maximumTsallisentropyaftertakingmanyiterations ofnegationnomatterwhatthevalueof q is,meanwhile, theprobabilitydistributionbecomesuniformdistribution.Hence,itisreasonabletochooseTsallisentropyin NPDafter q isdetermined.Fromtheabove,itcanbe knownthatthenegationcanconsumesomeinformationandincreasetheTsallisentropy.Finally,thearticle alsodiscussestheapplicationofnegationintargetrecognitionbasedonsensorfusion,whichshowsthat negationcanobtainamorereasonabledecisionwhen theconflictishigh.Insum,negationnotonlyprovides anewviewtoobtaininformationfromanotherside basedonknowninformation,butalsocaninclude impreciseclasstomakebetterdecision.Combiningthe knowninformationandtheinformationofnegation canincreasetheaccuracyofdecision-making.
Thisarticleisapreliminarystudytoobtainthe NPDbasedonuncertaintymeasurements.Thisworkis donemainlytodesignamoreefficientnegationprocess anduncertaintymeasurement,andexpandtheapplicationofnegation.Besides,itisessentialtodeterminethe uncertaintyrelatedtothenegationandstudyits properties.
Declarationofconflictinginterests
Theauthor(s)declarednopotentialconflictsofinterestwith respecttotheresearch,authorship,and/orpublicationofthis article.
Funding
Theauthor(s)disclosedreceiptofthefollowingfinancialsupportfortheresearch,authorship,and/orpublicationofthis
article:TheworkispartiallysupportedbyNationalNatural ScienceFoundationofChina(GrantNos61573290, 61503237).
ORCIDiD
YongDeng https://orcid.org/0000-0001-9286-2123
References
1.LiuC,GrenierD,JousselmeAL,etal.Reducingalgorithmcomplexityforcomputinganaggregateuncertainty measure. IEEETSystManCyb:PartASystHuman 2007;37(5):669–679.
2.HanYandDengY.AnevidentialfractalAHPtarget recognitionmethod. DefenceSciJ 2018;68(4):367–373.
3.LiuZG,PanQandDezertJ.Anewbelief-basedK-nearestneighborclassificationmethod. PatternRecogn 2013; 46(3):834–844.
4.MohamedAB,GunsekaranM,DoaaES,etal.IntegratingthewhalealgorithmwithTabusearchforquadratic assignmentproblem:anewapproachforlocatinghospitaldepartments. ApplSoftComput 2018;13:530–546.
5.ChenLandDengX.Amodifiedmethodforevaluating sustainabletransportsolutionsbasedonAHP andDempster–Shaferevidencetheory. ApplSci 2018; 8(4):563.
6.DengXandDengY.D-AHPmethodwithdifferent credibilityofinformation. SoftComput 2019;23(2): 683–691.
7.SeitiHandHafezalkotobA.Developingpessimisticoptimisticrisk-basedmethodsformulti-sensorfusion:an interval-valuedevidencetheoryapproach. ApplSoft Comput 2018;72:609–623.
8.WangY,ZhangKandDengY.Basebelieffunction:an efficientmethodofconflictmanagement. JAmbientIntel HumanComput 2018;2018:1–11.
9.AxenieC,RichterCandConradtJ.Aself-synthesis approachtoperceptuallearningformultisensoryfusion inrobotics. Sensors 2016;16:101751.
10.HanYandDengY.Anenhancedfuzzyevidential DEMATELmethodwithitsapplicationtoidentifycriticalsuccessfactors. SoftComput 2018;22(15):5073–5090.
11.SuX,MahadevanS,HanW,etal.Combiningdependent bodiesofevidence. ApplIntell 2016;44(3):634–644.
12.HanYandDengY.Anovelmatrixgamewithpayoffs ofmaxitivebeliefstructure. IntJIntellSyst 2019;34(4): 690–706.
13.Abdel-BassetM,MohamedMandSmarandacheF.An extensionofneutrosophicAHP-SWOTanalysisforstrategicplanninganddecision-making. Symmetry-Basel 2018;10(4):116.
14.LachaizeM,LeHgarat-MascleS,AldeaE,etal.Evidentialframeworkforerrorcorrectingoutputcodeclassification. EngApplArtifIntel 2018;73:10–21.
15.BehrouzMandAlimohammadiS.Uncertaintyanalysis offloodcontrolmeasuresincludingepistemicandaleatoryuncertainties:probabilitytheoryandevidencetheory. JHydrolEng 2018;23(8):04018033.
16.WuB,YanX,WangY,etal.AnevidentialreasoningbasedCREAMtohumanreliabilityanalysisinmaritime accidentprocess. RiskAnal 2017;37(10):1936–1957.
17.ChenLandDengY.Anewfailuremodeandeffects analysismodelusingDempster-Shaferevidencetheory andgreyrelationalprojectionmethod. EngApplArtif Intell 2018;76:13–20.
18.ChatterjeeK,PamucarDandZavadskasEK.Evaluating theperformanceofsuppliersbasedonusingtheR’AMATEL-MAIRCAmethodforgreensupplychainimplementationinelectronicsindustry. JCleanProd 2018;84: 101129.
19.GaoXandDengY.Thenegationofbasicprobability assignment. IEEEAccess 2019;PP(99):1932.
20.JanghorbaniAandMoradiMH.Fuzzyevidentialnetworkanditsapplicationasmedicalprognosisanddiagnosismodels. JBiomedInform 2017;72:96–107.
21.SeitiH,HafezalkotobA,NajafiS,etal.Arisk-based fuzzyevidentialframeworkforFMEAanalysisunder uncertainty:anintervalvaluedDSapproach. JIntell FuzzySyst 2018;35(2):1419–1143.
22.LiuZ,PanQ,DezertJ,etal.Combinationofclassifiers withoptimalweightbasedonevidentialreasoning. IEEE TFuzzySyst 2018;26(3):1217–1230.
23.SunRandDengY.Anewmethodtoidentifyincomplete frameofdiscernmentinevidencetheory. IEEEAccess 2019;7:15547–15555.
24.DengX,HuY,DengY,etal.Environmentalimpact assessmentbasedonDnumbers. ExpertSystAppl 2014; 41(2):635–643.
25.RosenbergE.Maximalentropycoveringsandtheinformationdimensionofacomplexnetwork. PhysLettA 2017;381(6):574–580.
26.SrivastavaAandMaheshwariS.Somenewpropertiesof negationofaprobabilitydistribution. IntJIntelSyst 2017;33(5):1133–1145.
27.YagerRR.Onthemaximumentropynegationofaprobabilitydistribution. IEEETFuzzySyst 2015;23(5): 1899–1902.
28.TsallisC.Nonadditiveentropy:theconceptanditsuse. EurPhysJA 2009;40(3):257–266.
29.TsallisC.PossiblegeneralizationofBoltzmann-Gibbs statistics. JStatPhys 1998;52(1–2):479–487.
30.TsallisC. Introductiontononextensivestatistical mechanics:approachingacomplexworld.NewYork: Springer,2009.
31.KangB,DengY,HewageK,etal.GeneratingZ-number basedonOWAweightsusingmaximumentropy. IntJ IntellSyst 2018;33(8):1745–1755.
32.PramanikS,DalapatiS,AlamS,etal.NC-crossentropy basedMADMstrategyinneutrosophiccubicsetenvironment. Mathematicas 2018;6(5):67.
33.XiaoF.Animprovedmethodforcombiningconflicting evidencesbasedonthesimilaritymeasureandbelieffunctionentropy. IntJFuzzySyst 2018;20(4):1256–1266.
34.JaynesET.GibbsvsBoltzmannentropies. AmJPhys 1965;33(5):391–398.
35.YagerRR.IntervalvaluedentropiesforDempster–Shaferstructures. KnowlBasedSyst 2018;161:390–397.
36.AtliBG,MicheY,KalliolaA,etal.Anomaly-based intrusiondetectionusingextremelearningmachineand aggregationofnetworktrafficstatisticsinprobability space. CognitiveComput 2018;10(5):848–863.
37.LiYandDengY.Generalizedorderedpropositions fusionbasedonbeliefentropy. IntJComputCommun Control 2018;13(5):792–807.
38.CetinkayaMC,UstunelS,OzbekH,etal.Convincing evidencefortheHalperin-Lubensky-MaeffectattheNSmAtransitioninalkyloxycyanobiphenylbinarymixturesviaahigh-resolutionbirefringencestudy. EurPhys JE 2018;41(10):129.
39.DongY,ZhangJ,LiZ,etal.Combinationofevidential sensorreportswithdistancefunctionandbeliefentropy infaultdiagnosis. IntJComputCommunControl 2019; 14(3):293–307.
40.YeJandCuiW.Exponentialentropyforsimplifiedneutrosophicsetsanditsapplicationindecisionmaking. Entropy 2018;20(5):357.
41.MardaniA,NilashiM,ZavadskasEK,etal.Decision makingmethodsbasedonfuzzyaggregationoperators: threedecadesreviewfrom1986to2017. IntJInform TechnolDecisMaking 2018;17(2):391–466.
42.JugadeSCandVictorinoAC.Gridbasedestimationof decisionuncertaintyofautonomousdrivingsystemsusing belieffunctiontheory. IFAC-PapersOnLine 2018;51(9): 261–266.
43.DuttaP.Modelingofvariabilityanduncertaintyin humanhealthriskassessment. MethodsX 2017;4:76–85.
44.PanLandDengY.Anewbeliefentropytomeasureuncertaintyofbasicprobabilityassignmentsbasedonbelieffunctionandplausibilityfunction. Entropy 2018;20(11):842.
45.DahooieJH,ZavadskasEK,AbolhasaniM,etal.A novelapproachforevaluationofprojectsusingan interval–valuedfuzzyadditiveratioassessment(ARAS) method:acasestudyofoilandgaswelldrillingprojects. Symmetry 2018;10(2):45.
46.YinL,DengXandDengY.Thenegationofabasic probabilityassignment. IEEETFuzzySyst 2019;27(1): 135–143.
47.ClausiusR. Themechanicaltheoryofheat:withitsapplicationstothesteam-engineandtothephysicalproperties ofbodies.Wakefield:JohnVanVoorst,1867.
48.PredaV,DeduSandGheorgheC.NewclassesofLorenz curvesbymaximizingTsallisentropyundermeanand Giniequalityandinequalityconstraints. PhysicaA 2015; 436:925–932.
49.YinLandDengY.Towarduncertaintyofweightednetworks:anentropy-basedmodel. PhysicaA 2018;508: 176–186.
50.DengWandDengY.Entropicmethodologyforentanglementmeasures. PhysicaA 2018;512:693–697.
51.DengXandJiangW.Dependenceassessmentinhuman reliabilityanalysisusinganevidentialnetworkapproach extendedbybeliefrulesanduncertaintymeasures. Annal NuclearEnerg 2018;117:183–193.
52.XiaoF.Multi-sensordatafusionbasedonthebelief divergencemeasureofevidencesandthebeliefentropy. InformFusion 2019;46:23–32.
53.ShannonCE.Amathematicaltheoryofcommunication. BellSystTechJ 1948;27(3):379–423.
54.KhammarAHandJahanshahiSMA.OnweightedcumulativeresidualTsallisentropyanditsdynamicversion. PhysicaA 2018;491:678–692.
55.KhammarAandJahanshahiS.QuantilebasedTsallis entropyinresiduallifetime. PhysicaA 2018;492:994–1006.
56.WilkGandWodarczykZ.ExampleofapossibleinterpretationofTsallisentropy. PhysicaA 2008;387(19):4809–4813.
57.LopesFM,OliveiraEADandCesarRMJr.Inferenceof generegulatorynetworksfromtimeseriesbyTsallis entropy. BMCSystBiol 2011;5(1):61–61.
58.BritoLRJD,SilvaSandTsallisC.Roleofdimensionality incomplexnetworks. SciReport 2016;6:27992.
59.KoikeCYandHigaCHA.Inferenceofgeneregulatory networksusingcoefficientofdetermination,Tsallis entropyandbiologicalpriorknowledge.In: Proceedings oftheIEEEinternationalconferenceonbioinformaticsand bioengineering,Taichung,31October–2November2016, pp.64–70.NewYork:IEEE.
60.HanYandDengY.Ahybridintelligentmodelfor assessmentofcriticalsuccessfactorsinhighriskemergencysystem. JAmbientIntelHumanComput 2018;9(6): 1933–1953.
61.HuynhVN,NguyenTTandLeCA.AdaptivelyentropybasedweightingclassifiersincombinationusingDempster–Shafertheoryforwordsensedisambiguation. ComputSpeechLang 2010;24(3):461–473.
62.KhanM,SonLH,AliM,etal.Systematicreviewofdecisionmakingalgorithmsinextendedneutrosophicsets. Symmetry-Basel 2018;10(8):314.
63.DuttaP.AnuncertaintymeasureandfusionruleforconflictevidencesofbigdataviaDempster–Shafertheory. IntJImageDataFusion 2018;9(2):152–169.
64.ZhangHandDengY.Enginefaultdiagnosisbasedon sensordatafusionconsideringinformationqualityand evidencetheory. AdvMechEng 2018;10(10):809184.
65.LiM,ZhangQandDengY.Evidentialidentificationof influentialnodesinnetworkofnetworks. ChaosSoliton Fract 2018;117:283–296.