Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

NDivisionofInformatics,UniversityofEdinburgh

DivisionofInformatics

http://www.77cn.com.cnrmatics.ed.ac.uk/

June2000

Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

ACase-BasedReasoningFrameworkforEnterprise

ModelBuilding,SharingandReusing

JessicaChen-Burger,DaveRobertson,JustineStaderInformaticsResearchReportEDI-INF-RR-0022DIVISIONofINFORMATICS

Arti cialIntelligenceApplicationsInstituteInstituteforRepresentationandReasoning

June2000

ECAI’2000

Abstract:

Enterprisemodeldevelopmentisessentiallyalabour-intensiveexercise.HumanexpertsdependheavilyonpriorexperiencewhentheyarebuildingnewmodelsmakingitanaturaldomaintoapplyCaseBasedReasoningtechniques.Throughtheprovisionofmodelbuildingknowledge,automatictestinganddesignguidancecanbeprovidedbyrule-basedfacilities.Exploringtheseopportunitiesrequiresusnotonlytodeterminewhichformsofknowledgearegenericandthereforere-usable,butalsohowthisknowledgecanbeusedtoprovideusefulmodelbuildingsupport.ThispaperpresentsourexperiencesinidentifyingandclassifyingtheknowledgewhichexistsinIBM’sBSDMBusinessModelsandapplyingAItechniques,CBRandRule-Basedreasoningtogetherwithasymbolicsimulator,toprovidemorecompletesupportthroughouttheenterprisemodeldevelopmentlifecycle.

Keywords:EnterpriseModelling,ModelDevelopmentLifeCycle,CaseBasedReasoning,BusinessModelling,Pro-cessModelling,KnowledgeManagement,BSDM,FormalMethod.Copyrightc2000byTheUniversityofEdinburgh.AllRightsReserved

TheauthorsandtheUniversityofEdinburghretaintherighttoreproduceandpublishthispaperfornon-commercialpurposes.

Permissionisgrantedforthisreporttobereproducedbyothersfornon-commercialpurposesaslongasthiscopy-rightnoticeisreprintedinfullinanyreproduction.Applicationstomakeotheruseofthematerialshouldbeaddressedinthe rstinstancetoCopyrightPermissions,DivisionofInformatics,TheUniversityofEdinburgh,80SouthBridge,EdinburghEH11HN,Scotland.

Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

ACase-BasedReasoningFrameworkforEnterprise

ModelBuilding,SharingandReusing

Yun-HehChen-BurgerandDavidRobertsonandJussiStader

Abstract.Enterprisemodeldevelopmentisessentiallyalabour-intensiveexercise.HumanexpertsdependheavilyonpriorexperiencewhentheyarebuildingnewmodelsmakingitanaturaldomaintoapplyCaseBasedReasoningtechniques.Throughtheprovisionofmodelbuildingknowledge,automatictestinganddesignguidancecanbeprovidedbyrule-basedfacilities.Exploringtheseopportunitiesrequiresusnotonlytodeterminewhichformsofknowledgearegenericandthereforere-usable,butalsohowthisknowledgecanbeusedtoprovideusefulmodelbuildingsupport.ThispaperpresentsourexperiencesinidentifyingandclassifyingtheknowledgewhichexistsinIBM’sBSDMBusinessModelsandapplyingAItechniques,CBRandRule-Basedreasoningtogetherwithasymbolicsimulator,toprovidemorecompletesupportthroughouttheenterprisemodeldevelopmentlifecycle.

Key-wordsEnterpriseModelling,ModelDevelopmentLifeCycle,CaseBasedReasoning,BusinessModelling,ProcessModelling,KnowledgeManagement,BSDM,FormalMethod.

1Introduction

ThemaintaskofBSDM’sBusinessModellingistoidentifytwocon-ceptualcomponents:entitiesanddependencies.Entitiesarethingsthatabusinessneedstomanageanddependenciesaretherelation-shipsbetweenthesethings.Certainkindsofscenariosorrelation-shipsbetweenentitiesarecommontomanybusinesses.Hence,onewouldexpectthatthecorrespondingBSDMBusinessModelmapsre ectthesecommonalities.

Inpractice,IBMprovidesacatalogueofsuchgenericentitymod-els[8]:someofthemarestandardandexamplemodelsfromthemethodandsomeofthemwerespeci callydevelopedforselectedindustries.Providedwiththesegenericmodels,BSDMpractition-ershelpclientsbuildtheirbusinessmodelbyusingthisinformationimplicitlyorexplicitly.ForBSDMconsultancy,King[9]suggestedthreepossiblewaysofre-usinggeneric/knownmodelswhenaddress-inganewproblemdomain.

Back-PocketApproach:theclientsaremadeawareoftheexistenceofthesegenericmodels,buttheyareonlyusedtosupportcon-sultancy.Theclientwillseelittleornoneofthegenericmodel.Aconsultantkeepsthesegenericmodelsatthebackofhis/hermindandtailorsthemtotheclients’specialrequirements.

Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

formalisedtoprovidecoherentandcomprehensivesupportthrough-outthemodeldevelopmentlifecycle.Itconsiderstwoissues:issuchknowledgegenericandreusable,andhowcanthisknowledgebeusedtoprovideautomaticsupport.Thepaper rstdescribeshowCaseBasedReasoningtechniquescanbeusedtoprovideacommonplatformforknowledgesharing.Itthenpresentstowhichextentthisknowledgecanbeformalisedandprovideassistanceformodelbuild-ingactivities.

2TheModellingSupportFramework

of a business model development.

Figure1.ArchitectureofGenericModelAdvisor

Figure1showsthemodellingframeworkwhichprovidesauto-maticfacilitiestosupporttheiterativeplan-build-test-re nemodel-lingdevelopmentlifecycleasshowninFigure2.

Figure2.ThePlan-Build-Test-Re nedevelopmentcycle

Twointegratedknowledgebasedsupporttools,GenericModelAdvisor(GMA)andKnowledgeBasedSupportToolforBusinessModelling(KBST-BM),havebeenbuilt.SinceaBSDM’sbusinessmodelisorganisedandpresentedinviewsanddiagrams,thesearethe”units”thatGMAstoresandretrieves.GMAidenti esandassignsindices(featureswhichcharacteriseamodel)totheproblem,i.e.theuser-de nedBSDMmodel.Theseindices,togetherwiththeembed-deddomainknowledge,inourcasetheEntityConceptualHierarchyandMatchRules,arepassedtothepatternmatchingalgorithmwhichcomparestheindicesoftheproblemandthoseofthegenericmodels

2

intheGenericModelLibrarytoretrieveasetofreferencemodelswhichexhibitsimilarcharacteristicstotheinputmodel.

Atthisstagetheretrievedsimilargenericmodelsarenotyetex-aminedtodeterminewhichisabettermatchforthecurrentproblem.Forsuchacomparison,GMAprovidesa exibleSimilarityAssess-mentFunctionwhichenablesthedeploymentofabuilt-inheuristicmethodortheuserscandynamicallymakeuptheirownevaluationmethodtoexplorespeci cmatchesbasedontheidenti edindicesofthemodel.

Thebestmatchingcase,accordingtothechosensimilarityassess-mentmethodandananalysisreportofsimilaritiesanddifferencesbetweentheusermodelandtheretrievedreferencemodeltogetherwithsuggestionsabouthowtoeliminatethecausesofthedifferences,aregiventotheuser.Theusercanthenreadthereportand/oraskthesystemtopresentadifferentmatchingresultforanothergenericmodel.Matchesareshowninthedescendingorderoftheirscoresinthechosensimilarityassessmentmethod.Asummaryofallofthematchesshowntotheuserisproducedseparatelywhichrecordsthesimilaritymeasurementsofeachmatchtogivetheuseranoverviewofallpossiblemappingsandtoallowrevisitingofselectedgenericmodels.

Auser-de nedmodelmaybematchedwithmorethanonegenericmodels.Theusercanchoosetomodifyhis/hermodelandrepeattheabovemodellingcycleasapartofaniterativeprocess.Iftheuserhasdecidedtousethereferencemodelasabasistogenerateanewmodel,theusercanexportthechosenreferencemodelfromthelib-rary.Atanystageofthemodeldevelopment,theusercanchoosetousetheveri cationandvalidationfacilitiesprovidedbyKBST-BMtocheckforthecompleteness,soundnessandappropriatenessofthebuiltmodel.

Whentheuserissuf cientlysatis edwithhis/hermodel,he/shecanretainthisnewmodel,i.e.writeitbacktoGMA,by rstlygen-eralisingthenewmodel,verifyingandvalidatingthegeneralisedmodelusingtheintegratedtoolKBST-BM,andthenstoringthenewgenericmodelbacktotheGenericModelLibrary.TheCaseBasedReasoningCycleisnowcompleted,andGMA’sknowledgecanbeenrichedandevolvedthroughtimeviatheinclusionofnewlyac-quiredknowledgeduringoperations.GMAdoesnotprovideanauto-maticadaptationfacilityfortworeasons.First,thereisnoabsolutestandardwhich tsallbusinessesindeterminingwhetherornotapar-ticulardesignisthemostappropriateoneforabusiness.Secondly,althoughcommonpracticesaresharedbymanybusinesses,businessmodelsareingeneralorganisation-dependentandbuildingagoodmodelrequiresunderstandingoftheorganisation’soperationandaconsensuswithintheorganisationwhichmaynotbeavailableorformalisableduetothesizeandnatureoftherequiredknowledge[4].Bothissueshavetoberesolvedbeforehighqualityautomaticadapt-ationcanbeprovided.

TheinnerKBST-BMsystemboxintheFigure1illustrateshowKBST-BMcanassistincompletingtheCBRcycle.Itprovidesanin-dependentveri cationandvalidation(V&V)facilities(fromtheuser)andisincludedinthe“Test”activityinthestandardmodeldevelop-mentprocessshowninFigure2.ThisV&Vapproachandimplement-ationdetailsofKBST-BMaregivenin[3].

3Indexing,MatchingandSimilarityAssessment

Indicesarefeatureswhichcanbeusedtodistinguishmodelsinthecasememoryandto ndappropriatematchesbetweenagivenprob-lemandpreviousmodels.InthecontextofaBSDMbusinessmodel,thesedistinguishingcharacteristicsareembeddedinthesemanticsof

Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

entities,thearchitectureofabusinessmodel,andthebusinessareathatamodeldescribes.

Simplycomparingthegraphicalrepresentationofbusinessmod-elsisnotsuf cient.Forexample,drawinganexistingmodelupside-downdoesnotmakeitadifferentmodel,thesemanticsoftheinter-relationships(dependencies)betweenentitiesmustbetakenintoaccount.Furthermore,businesscontextualsimilaritiesmaybedis-guised.Forinstance,ifabusinessmodelisamoreelaboratedorspe-cialisedversionofanotherone(orviceversa),thenthesetwomodelsnormallywillnothavethesamearchitecture(e.g.onemayexpandpartsofthemodelinsomeareas),andoftentheydonotsharethesameentities(http://www.77cn.com.cningdomainspeci cvocabulariesinstead).How-ever,becausetheyareessentiallydescribingthesimilarbusinessop-erations,itwillbeusefultoreferonetotheother.

TobeabletomakemeaningfulcomparisonsbetweenBSDMmod-els,onemusthaveanintegralunderstandingofthebusinesscontextwhichisdescribedinboththearchitectureofamodelaswellasthebusinesscontextthateachentityrepresents.Wecapturepartofthiscontextthroughtypingofentitiesviaaconcepthierarchy.

Typesprovideastructuretoallocateconceptualcategoriesandnor-mallydescribemore“general”concepts.ConcreteEntitiespresentmorespecialisedconceptsandincludeentitieswhichareusedinrealbusinessmodels(asopposedtoageneralisedmodel).AnarrowfromentityBtoentityAindicatesanis-arelationshipfromBtoA,i.e.Bis-aA.

TheEntityConceptualHierarchycapturesthesemanticsofalloftheentities(intheuserandreferencemodels)aswellastherelation-shipsbetweenthemanditcanbeusedtoidentifyandmatchsimilarentitiesusedintheuserandreferencemodels.

3.2CaseRetrievingandSimilarityAssessment

ThePatternMatchingAlgorithmcomparesthecontextualandarchi-tectureinformationofthegivenusermodelwiththatofalloftheref-erencemodelsstoredintheGenericModelLibrary.Severaltypesofinformationistakenintoaccount.Dothesemodelsdescribeasimilarbusinessarea?Aretheycapturingsimilarconcepts?Dotheyfollowsimilarbusinessrules?Thecontextualandarchitectureinformationisstoredinthebusinessarea,view,links,dependencies,andintheentities.

ProvidedwithknowledgeembeddedinECH,onecannowmatchviews,dependenciesandentitiestodetermineiftwodifferentmod-elsaresuf cientlysimilar.Tomatchentities,forinstance,entitieswhichhavethesamenameinbothuserandreferencemodelspro-duceapositivematch.However,similarbutvariantentities(siblingrelationshipsintheECH),or“stream-line”specialisations(e.g.par-entandchild,orgrandparentandgrandchildrelationships)mayalsoproduceapositivematch.Whendecidingwhichisabettermatchbetweenentities,theclosertherelationshipisbetweenthetwoentit-iesontheECH,thebetterqualityofamatchitis.

Ausermodelmayincludeseveralgenericmodels.Ontheotherhand,agenericmodelmayincludeorpartiallyoverlapwiththeusermodel.Figure4showsthepossibilitieshowausermodelmaybemappedtoagenericmodel.

U

G

UCASE II

G

U

GCASE III

3.1EntityConceptualHierarchy(ECH)

BSDMprovidesEntityFamilieswhichprovideentitiesingroupsac-cordingtowhereandhowtheycanbeusedinabusinessmodel.BSDMmodellersuseEntityFamiliesasastartingpointwhentryingtoidentifyentitiesforanewmodel.Theyalsouseitasaguidelinetocheckthearchitectureofthemodel.WeorganiseinformationgivenintheEntityFamiliesinataxonomichierarchy,calledtheEntityCon-ceptualHierarchy

.

CASE IEquivalentG

User model is included in the Generic ModelG

UU

CASE IVCASE V

User model is not included in the generic model, but the generic model is fully included in the User Model

U

G

G

U

U

CASE VIII

G

CASE VICASE VII

User model is partially overlapping with the generic model

Figure4.

Figure3.

APartofEntityConceptualHierarchy(ECH)

PossibleMatchingbetweenUserModelsandGenericModels

Figure3showsascreenshotfromGMAwhichcapturesapartoftheEntityConceptualHierarchywhichcontainsthesuggesteden-titiesforthetoplayer(layer1)ofaBSDMbusinessmodel.Twotypesofclasseshavebeenusedtodescribeentities:theshadedrect-angularboxesrepresenttheAbstractEntityTypes,andtheclearrect-angularboxesrepresenttheConcreteEntityTypes.AbstractEntity

3

Asouraimistoseekforthebestorbettermatch,naturallya100%matchisalwaysgiventhehighestpriority,thereforeCASEI.Thesecondpreferencegoestoamatchinwhichanusermodelisfullyincludedbytheselectedgenericmodel,henceCASEIIandIII.How-ever,CASEIIissuperiorthanCASEIIIbecauseitsgenericmodelismoresimilartotheusermodel:ithasasmallerdifferencecomparedbetweenthetwomodels.

Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

Whenausermodelisnotfullycovered,wepreferamatchwheretheusermodelhasabettercoveragefromtheselectedgenericmodel,henceCASEIVissuperiortoCASEV,VIandVIIwhichareallmoresuperiorthanCASEVIII.InthecaseofV,VIandVIIwherethecoverageofcommonalityoftheusermodelaresimilar,thequal-ityofthematchedgenericmodelshouldbetakenintoaccount,i.e.agenericmodelwhichismoresimilartotheusermodeshouldbepre-ferred.SincethegenericmodelinCASEVisentirelyincludedintheusermodel,itisthemostsimilar(orrelevant)onetotheusermodel,CASEVIisinsecondplace,andCASEVIIistheleastsimilaronetotheusermodelsinceithasacomparativelysmallercommonportionwiththeusermodel.

Basedonourpreferences, vediscriminatingcriteriaareidenti- ed:thematchingresultofthecapturedbusinessareas,thematch-ingratiooflinks(dependencies)intheselectedreferencemodel,thematchingratioofentitiesintheselectedreferencemodel,thematch-ingratiooflinks(dependencies)intheusermodelandthematchingratioofentitiesintheusermodel.

HEURISTIC SIMILARITY ASSESSMENT FUNCTIONGiven two matches, X and Y

if match-view(X) > match-view(Y) then SELECT Xelse if match-view(X) = match-view(Y)) and

match-data-link(X) > match-data-link(Y) then SELECT Xelse if match-view(X) = match-view(Y)) and

match-data-link(X) = match-data-link(Y) and

match-data-entity(X) > match-data-entity(Y) then SELECT Xelse if match-view(X) = match-view(Y)) and

match-data-link(X) = match-data-link(Y) andmatch-data-entity(X) = match-data-entity(Y) and

match-case-link(X) > match-case-link(Y) then SELECT Xelse if match-view(X) = match-view(Y)) and

match-data-link(X) = match-data-link(Y) andmatch-data-entity(X) = match-data-entity(Y) andmatch-case-link(X) = match-case-link(Y) and

match-case-entity(X) > match-case-entity(Y) then SELECT Xelse SELECT Y

Figure5.TheHeuristicSimilarityEvaluationFunction

Figure5showstheheuristicevaluationmethodprovidedbyGMA.Itprovidesameanstousetheevaluationcriteriainselectingabet-termodelwhichcomplieswiththepreferenceorderdemonstratedearlier.Thismethodproducesgoodresultsusingourtestdata(seeSection4).Alternatively,theusercandynamicallydesigntheirownevaluationmethodsusingWeightedCity-Blockevaluationfunctionbasedontheabovecriteria,iftheywishtoexplorespeci caspectsofmodelsinthelibrary.

4Evaluation

Forevaluationpurposes,weobtainedavarietyofBSDMmodelsfromdifferentdomains.Partofarealindustrialmodelwhichwasde-velopedbyaninternationalautomobilecompany.3Onegenericbusi-nessmodelforsmallandmedium-sizedrestaurantwasdevelopedbasedoninterviewsofthreeindependentfamilyrestaurant(ex-)ownerstoenlargeourtestingbase.WealsocapturedexampleandstandardmodelsfromBSDMandstoredtheminourGenericModel

Enterprise model development is essentially a labour-intensive exercise. Human experts depend heavily on prior experience when they are building new models making it a natural domain to apply Case Based Reasoning techniques. Through the provision of model

KBST-BMintegrateswithGMAtogetherprovideanadequateframeworkforCBR,i.e.automaticindexinginputdata,retrievingrelevantcasesfromlibrary,comparingandanalysinginputwithse-lectedcases,revisingcasesforcurrentproblem,verifyingandval-idatinginput,andretainingthenewinputsforfuturereference.ThisallowsustousethelargerKBST-BMBSDMmodellingenvironmentintheadaptationphaseoftheCBRcycle.Wetestedthisrouteusingtheautomobileandrestaurantmodels.

[14]E.StrouliaandA.K.Goel,‘GenericTeleologicalMechanismsand

TheirUseinCaseAdaptation’,ProceedingsoftheFourteenthAnnualConferenceoftheCognitiveScience,(1992).Northvale,N.J.,Erlbaum.[15]K.Sycara,R.Guttal,J.Koning,S.Narasimhan,andD.Navin-chandra,‘CADET:ACase-basedSynthesisToolforEngineeringDesign’,InternationalJournalforExpertSystems,4(2),pp.157–188,(1992).http://www.cs.cmu.edu/afs/cs.cmu.edu/project/cadet/ftp/docs/CADET.html.

5Conclusion

Successfulbusinessmodeldevelopmentrequiresbothmethodolo-gicalandapplicationdomainknowledgeandexperience.Unfortu-nately,fewpeoplepossessallofthesecapabilities.OurstudiesofapplyingCBRandRule-Basedtechniqueswhicharebasedonaco-herentunderlyingformalmethodshowshowmodelbuildingknow-ledgecanbeobtained,reusedandusedtoprovideautomaticveri c-ationandvalidationfacilities.Webelievethatwiththissupportweareabletoenhancethelevelofknowledgesharing,andabilityofproblemsolving.Moreimportantly,itaddstoourunderstandingofhowthissortofseeminglyinformalmethodcan tintopartsofthedesignlifecyclewhichrequireformalmodels.

REFERENCES

[1]Klaus-DieterAlthoff,EricAuriol,RalphBarletta,andMichelManago,

AnAIPerspectivesReport:AReviewofIndustrialCase-BasedReas-oningTools,AnAIPerspectiveReport,AIIntelligence,P.O.Box95,OxfordOX27XL,1995.

[2]J.Barber,S.Bhatta,A.Goel,M.Jacobsen,M.Pearce,L.Penberthy,

M.Shankar,andE.Stroulia,IntegratingCase-BasedReasoningAndMultimediaTechnologiesForInterfaceDesignSupport,InArti cialIn-telligenceInDesign,Editor:J.G.Boston,KluwerAcademicPublisher,1992.

[3]Yun-HehChen-Burger,DaveRobertson,andJussiStader,‘Knowledge-BasedAutomaticVeri cationandValidationforBusinessMod-els’,DARPA-JFACCSymposiumonAdvancesinEnterpriseControl,(November1999).

[4]Yun-HehChen-Burger,DavidRobertson,andJussiStader,‘Formal

SupportforanInformalBusinessModellingMethod’,TheSpecialIssueforTheInternationalJournalofSoftwareEngineeringandKnowledgeEngineering,(February2000).

[5]E.Domeshek,J.Kolodner,andC.Zimring,‘TheDesignofaToolKit

forCase-BasedDesignAids’,ProceedingsoftheThirdInternationalConferenceonArti cialIntelligenceinDesign,(1994).

[6]B.Faltings,CaseReuseByModel-BasedInterpretation,inIssuesand

ApplicationsofCase-BasedReasoninginDesign,Editor:M.L.Maher,P.Pu,LawrenceErlbaumAssociates,Hillsdale,N.J.,1997.pp.30-60.[7]T.R.Hinrichs,‘TowardsanArchitectureforOpenWorldProblemSolv-ing’,ProceedingsofCBRworkshop,pp.182–189,(1988).MorganKaufmann,SanFrancisco.

[8]IBMUnitedKingdomLimited,389ChriswickHighRoad,LondonW4

4AL,England,BusinessSystemDevelopmentMethod:BusinessMap-pingPart1:Entities,2ndedn.,May1992.

[9]MartinKing,‘KnowledgeReuseinBusinessDomainsExperiencewith

IBMBSDM’,Technicalreport,Arti cialIntelligenceApplicationInsti-tute,(1995).

[10]JanetKolodner,Case-BasedReasoning,MorganKaufmannPublishers,

Inc.,2929CampusDrive,suite260,SanMateo,CA,USA,1993.

[11]M.L.Maher,B.Balachandran,andD.M.Zhang,Case-BasedReason-ingInDesign,LawrenceErlbaum,1995.

[12]M.L.MaherandA.GomezdeSilvaGarza,‘DevelopingCase-Based

ReasoningforStructuralDesign’,IEEEExpert,IntelligentSystemsandTheirApplications,11(3),(June1996).

[13]S.Narashiman,K.Sycara,andD.Navin-Chandra,Representationand

SynthesisofNon-MonotonicMechanicalDevices,InIssuesandApplic-ationsofCase-BasedReasoninginDesign,Editor:M.L.Maher,P.Pu,LawrenceErlbaumAssociates,Hillsdale,N.J.,1997.

5