+ All Categories
Home > Documents > 39248830 Expert Systems

39248830 Expert Systems

Date post: 10-Apr-2018
Category:
Upload: manish-baranwal
View: 216 times
Download: 0 times
Share this document with a friend

of 82

Transcript
  • 8/8/2019 39248830 Expert Systems

    1/82

    EXPERTEXPERTSYSTEMSYSTEM

  • 8/8/2019 39248830 Expert Systems

    2/82

    AnAn expertexpert systemsystem containscontains knowledgeknowledge

    aboutabout aa particular particular fieldfield toto assistassist humanhuman

    expertsexperts oror provideprovide informationinformation toto peoplepeople

    whowho dodo notnot havehave accessaccess toto anan expertexpert inin

    thethe particularparticular fieldfield..

  • 8/8/2019 39248830 Expert Systems

    3/82

    What is an Expert System ?What is an Expert System ?

    AnAn expertexpert systemsystem isis defineddefined inin differentdifferent waysways::

    SoftwareSoftware program program designeddesigned toto replicatereplicate thethedecisiondecision makingmaking processprocess ofof aa humanhuman expertexpert

    aa program program thatthat simulatessimulates thethe judgment judgment andand

    behavior behavior ofof humanhuman thatthat hashas knowledgeknowledge andand

    experienceexperience inin aa particularparticular fieldfield

    aa softwaresoftware thatthat givesgives computercomputer humanhuman likelike

    abilitiesabilities..

  • 8/8/2019 39248830 Expert Systems

    4/82

    Putting all these definitions together,expertPutting all these definitions together,expert

    system can be defined as :system can be defined as :

    InteractiveInteractive computercomputer--basedbased systemssystems thatthat respondrespond totothethe questions,questions, asksasks forfor clarifications,clarifications, makesmakes

    recommendations,recommendations, andand helpshelps thethe useruser makemake aa

    betterbetter andand reasonablereasonable decisiondecision..

  • 8/8/2019 39248830 Expert Systems

    5/82

    ExpertExpert SystemSystem (ES)(ES) derivesderives itsits knowledgeknowledgefromfrom aa realreal livelive domaindomain expert,expert, aa humanhuman expertexpert ininparticular particular fieldfield.. AA user,user, andand expertexpert oror aaknowledgeableknowledgeable engineerengineer designsdesigns itit..

    TheThe systemsystem designerdesigner interviewsinterviews andand observesobservesthethe expertexpert whowho maymay be be aa technician,technician, airbusairbusengineer,engineer, etc,etc, andand finallyfinally convertsconverts hishis wordswords andand

    actionsactions andand yearsyears ofof experienceexperience intointo aa factualfactualknowledgeknowledge andand rulesrules oror simplysimply aa knowledgeknowledge basebase..

  • 8/8/2019 39248830 Expert Systems

    6/82

    KnowledgeKnowledge BaseBase isis aa collectioncollection ofof informationinformation

    thatthat containscontains factsfacts andand aa systemsystem ofof rulesrules forfor

    determiningdetermining thethe relationshiprelationship betweenbetween thosethose factsfacts..

    TheThe termsterms expertexpert systemsystem andand knowledgeknowledge basebase areare

    oftenoften interchangeablyinterchangeably usedused..

    ExpertExpert SystemSystem goesgoes throughthrough aa numbernumber ofof

    possible possible waysways andand atat lastlast comescomes upup withwith anan

    acceptableacceptable solutionsolution toto thethe problemproblem..

  • 8/8/2019 39248830 Expert Systems

    7/82

    ItIt appliesapplies preset preset rulesrules forfor determiningdetermining thethe

    problemproblem.. TheThe rulesrules cancan rangerange fromfrom 100100 toto 100100,,000000

    dependingdepending onon thethe complexitycomplexity ofof thethe problemproblem..

    ForFor exampleexample::

    IfIf

    LightLight--dodo--notnot--workwork

    ThenThen

    CheckCheck--thethe--batterybattery

  • 8/8/2019 39248830 Expert Systems

    8/82

    InIn casecase thethe batterybattery isis workingworking properlyproperly thenthenwewe cancan applyapply anotheranotherIFIF--ThenThen procedureprocedure toto havehave

    thethe desireddesired resultresult

    IfIf

    Battery not deadBattery not dead

    ThenThen

    CheckCheck--thethe--connectorsconnectors

  • 8/8/2019 39248830 Expert Systems

    9/82

    ToTo varyingvarying degrees,degrees, expertexpert systemssystems cancanreasonreason andand makemake decisionsdecisions.. RememberRemember DeepDeep

    BlueBlue aa computercomputer whichwhich defeateddefeated humanhuman chesschess

    master,master, thethe systemsystem containedcontained thousandsthousands ofof

    possibilitiespossibilities ofof playingplaying chesschess trickstricks..

  • 8/8/2019 39248830 Expert Systems

    10/82

    Expert system dialogue on skin diseaseExpert system dialogue on skin disease

    diagnosis with a patientdiagnosis with a patient

    ProgramProgram :: Are there any spot on the bodyAre there any spot on the body

    UserUser :: YesYes

    ProgramProgram :: What color spotsWhat color spots

    UserUser :: Reddish brownReddish brown

    ProgramProgram :: Are there more spots on theAre there more spots on the

    trunk than elsewheretrunk than elsewhere

    UserUser :: NoNo

    ProgramProgram :: The patient has measlesThe patient has measles

    This show expert system can produce bizarre output.This show expert system can produce bizarre output.

  • 8/8/2019 39248830 Expert Systems

    11/82

    Comparison ofHuman Expertise with ExpertComparison ofHuman Expertise with Expert

    systemsystem

    HumanHuman expertiseexpertise isis perishableperishable:: HumanHuman maymaychangechange jobs, jobs, theythey cancan become become ill,ill, maymay bebeinjured/deadinjured/dead andand soso onon.. ComputerComputer expertiseexpertise isispermanentpermanent..

    HumanHuman expertiseexpertise isis notnot alwaysalways consistentconsistent::HumanHuman expertexpert cancan havehave offoff days,days, busybusy schedulesschedules

    andand soso onon.. ComputersComputers areare alwaysalways consistent,consistent, andandgivengiven identicalidentical conditionsconditions theythey willwill alwaysalways givegivethethe samesame adviceadvice..

  • 8/8/2019 39248830 Expert Systems

    12/82

    HumanHuman expertiseexpertise isis difficultdifficult toto transfertransfer:: HumanHuman

    beingbeing cannotcannot bebe inin twotwo placesplaces atat thethe samesame timetime..ComputerComputer expertiseexpertise isis relativerelative easyeasy toto transfertransfer..ForFor example,example, anan expertexpert systemsystem runningrunning onon aa PCPCcouldcould bebe copiedcopied toto anotheranother PCPC atat aa differentdifferent site,site,

    oror downloadeddownloaded ontoonto aa networknetwork ofof PC,PC, oror evenevendownloadeddownloaded fromfrom InternetInternet..

    HumanHuman expertiseexpertise isis expensiveexpensive.. EmployeeEmployeesalariessalaries areare farfar greatergreater thanthan thethe costcost ofof PCPC

    hardwarehardware.. ExpertExpert system,system, eveneven takingtaking intointoaccountaccount developmentdevelopment cost,cost, areare muchmuch moremoreaffordableaffordable..

  • 8/8/2019 39248830 Expert Systems

    13/82

    HumanHuman areare creative,creative, oftenoften inspiredinspired;; ComputersComputersareare uninspireduninspired..

    HumanHuman areare flexibleflexible;; TheyThey areare easilyeasily adaptadapt ororintegrateintegrate theirtheir expertiseexpertise withwith otherother domaindomainknowledgeknowledge;; ComputersComputers areare notnot veryvery flexibleflexible ininthisthis respectrespect..

    HumanHuman posesposes commoncommon sensesense.. ExpertExpert systemsystemcannotcannot applyapply knowledgeknowledge toto aa problemproblem beyondbeyond thethedomaindomain knowledge,knowledge, andand thusthus havehave aa narrownarrow focusfocusofof thethe problemproblem..

    ConsistencyConsistency:: ThereThere isis alwaysalways aa consistencyconsistency ininthethe expertexpert systemsystem.. ThereThere willwill bebe nono ambiguityambiguity ororunderstandingunderstanding problemproblem toto understandunderstand thethe expertexpertsystemsystem..

  • 8/8/2019 39248830 Expert Systems

    14/82

    Organizational Benefits OfExpert SystemOrganizational Benefits OfExpert System

    KnowledgeKnowledge retentionretention:: TheThe knowledgeknowledge isis

    permanent,permanent, unlikeunlike thatthat ofof aa humanhuman expert,expert,

    whowho maymay switchswitch employment,employment, retire,retire, etcetc.. KnowledgeKnowledge distributiondistribution:: TheThe knowledgeknowledge

    cancan bebe distributeddistributed withwith inin aa companycompany sitesite oror

    anyany otherother sitesite anywhereanywhere inin thethe world,world, bybyusingusing networknetwork oror justjust duplicationduplication thethe expertexpert

    systemsystem onon standalonestandalone hardwarehardware..

  • 8/8/2019 39248830 Expert Systems

    15/82

    TrainingTraining:: TheThe explanationexplanation capabilitiescapabilities ofof expertexpert

    systemsystem areare suchsuch thatthat theythey oftenoften functionfunction asastrainingtraining system,system, useruser cancan seesee thethe chainchain ofof

    reasoningreasoning underlyingunderlying theirtheir decisionsdecisions andand hencehence

    gaingain aa better better understandingunderstanding ofof thethe problemproblem

    domaindomain..

    CompetitiveCompetitive edgeedge:: ExpertExpert systemsystem willwill oftenoften givegive

    aa companycompany aa competitivecompetitive edge,edge, owingowing toto highlyhighly

    increasedincreased responseresponse speeds,speeds, accuracyaccuracy ofof decisions,decisions,andand soso onon..

  • 8/8/2019 39248830 Expert Systems

    16/82

    CostCost reductionreduction:: TheThe costcost ofof providingproviding

    expertiseexpertise perper useruser isis lower,lower, computerscomputers dodo notnot

    paid paid largelarge salariessalaries forfor sharingsharing theirtheir

    knowledge,knowledge, oror consumeconsume significantsignificant resourceresourceapartapart fromfrom energyenergy andand maintenancemaintenance costscosts..

  • 8/8/2019 39248830 Expert Systems

    17/82

    Advantages to the userAdvantages to the user

    KnowledgeKnowledge accessibilityaccessibility:: TheThe expertiseexpertise isis

    alwaysalways availableavailable onon anyany computercomputer hardwarehardware..

    TrainingTraining:: TheThe valuevalue ofof trainingtraining cancan be be anan

    advantageadvantage toto bothboth employeremployer andand employeeemployee..

  • 8/8/2019 39248830 Expert Systems

    18/82

    ConsistencyConsistency:: TheThe useruser willwill knowknow thatthat adviceadvicegivengiven isis notnot subjectsubject toto humanhuman fallibilityfallibility.. TheThe

    expertexpert systemsystem isis notnot likelylikely toto havehave offoff--daysdays oror

    feelfeel sicksick oror havehave tootoo manymany otherother thingsthings onon itsits

    mindmind..

  • 8/8/2019 39248830 Expert Systems

    19/82

    BASICBASIC CONCEPTSCONCEPTS OFOF EXPERTEXPERT SYSTEMSYSTEM

    AnAn expertexpert systemsystem isis aa programprogram thatthat attemptsattempts totomimicmimic humanhuman expertiseexpertise by by applyingapplying inferenceinferencemethodsmethods toto aa specificspecific bodybody ofof knowledgeknowledge..

    DomainDomain :: bodybody ofof knowledgeknowledge isis callcall domaindomainDataData :: GroupGroup ofof alphabeticalphabetic symbolsymbol

    a)a) 55

    b)b) 66..00

    c)c) 88d)d) OasisOasis

    AA toto dd isis aa groupgroup ofof symbols,symbols, whichwhich havehave nono meaningmeaning onontheirtheir ownown..

  • 8/8/2019 39248830 Expert Systems

    20/82

    InformationInformation:: WhenWhen somesome contextcontext isis addedadded toto

    thethe symbol,symbol, thatthat havehave anyany benefitbenefitisis calledcalled informationinformation..

    a)a) $$55

    b)b) 66..00 marksmarks awardedawarded inin thethe competitioncompetition

    c)c) 88 degreedegreed)d) OasisOasis isis aa namename ofof rockrock groupgroup

  • 8/8/2019 39248830 Expert Systems

    21/82

    KnowledgeKnowledge:: KnowledgeKnowledge differsdiffers fromfrominformationinformation because because informationinformation

    isis passivepassive inin thethe sensesense thatthat itit doesdoes

    notnot givegive riserise toto furtherfurther generationsgenerations

    ofof informationinformation..

  • 8/8/2019 39248830 Expert Systems

    22/82

    Knowledge,Knowledge, conversely,conversely, isis activeactive inin thatthat itit cancan

    givegive riserise toto furtherfurther generationsgenerations ofof informationinformation..

    a)a) It is raining, therefore I shall get wet!It is raining, therefore I shall get wet!

    b)b) The temperature isThe temperature is 8 degree, so it is cold!8 degree, so it is cold!

    c)c) The engine of the car is hot; it must have beenThe engine of the car is hot; it must have been

    recently used!recently used!d)d) Oasis are affluent musicians.Oasis are affluent musicians.

  • 8/8/2019 39248830 Expert Systems

    23/82

    Basic Definitions :Basic Definitions :

    AlgorithmsAlgorithms:: AlgorithmsAlgorithms isis aa stepstep--byby--stepstep

    procedureprocedure thatthat solvessolves aa categorycategory

    ofof problemsproblems.. ForForExampleExample::processing processing ofof aa companycompany payrollpayroll

    useuse algorithms,algorithms, usingusing stepstep--byby--stepstep

    procedureprocedure onon thethe inputinput datadata suchsuch

    asas employeeemployee hours,hours, worked,worked,

    overtimeovertime raterate etcetc..

  • 8/8/2019 39248830 Expert Systems

    24/82

    HeuristicsHeuristics:: HeuristicHeuristic solvesolve aa problemproblem byby trialtrial

    andand errorerror guidedguided byby somesome

    referencereference toto aa predeterminedpredetermined goalgoal..

    ExampleExample:: ParkingParking system,system, RR && DD

    systemsystem etcetc....

    InferenceInference:: InferenceInference isis thethe drawingdrawing ofof

    conclusionconclusion fromfrom premisespremises..

  • 8/8/2019 39248830 Expert Systems

    25/82

    SearchSearch:: TheThe processprocess ofof startingstarting inin somesome initialinitial

    statestate andand attemptingattempting toto reachreach aa goalgoal statestatebyby evaluatingevaluating possiblepossible alternativealternativesolutionsolution..

    Space Search:Space Search:

    All of the possible states thatAll of the possible states thatcould be evaluated during acould be evaluated during asearch. The search space is oftensearch. The search space is oftenrepresented as an inverted tree,represented as an inverted tree,called a search tree.called a search tree.

  • 8/8/2019 39248830 Expert Systems

    26/82

    Blind search:Blind search: Blind search does not make anyBlind search does not make any

    use of the knowledge about theuse of the knowledge about the

    problem to guide the search.problem to guide the search.

    Heuristic search:Heuristic search: Heuristic search use domainHeuristic search use domain--specific knowledge to traverse thespecific knowledge to traverse the

    search spacesearch space

  • 8/8/2019 39248830 Expert Systems

    27/82

    COMPONENTSOF ANEXPERTSYSTEMCOMPONENTSOF ANEXPERTSYSTEM

    a). Knowledge Basea). Knowledge Base

    TheThe componentscomponents ofof expertexpert systemsystem thatthat containscontainsthethe systems systems knowledgeknowledge isis calledcalled knowledgeknowledge

    basebase systemsystem..

    ThisThis knowledgeknowledge isis frequentlyfrequently representedrepresented inin aacomputercomputer inin thethe form form ofof rulesrules:: theythey storestore thethe

    heuristicheuristic thatthat guideguide thethe humanhuman expertexpert..

  • 8/8/2019 39248830 Expert Systems

    28/82

    Example : A typical rule used by a house mortgageExample : A typical rule used by a house mortgage

    expert might be:expert might be:IF the applicant likes the houseIF the applicant likes the house

    AND the house is worth a surveyAND the house is worth a survey

    THEN advise the applicant to apply for a loanTHEN advise the applicant to apply for a loan

    Rule would usually be expressed in the formRule would usually be expressed in the form

    IF condition (s)IF condition (s)

    THEN actions (s)THEN actions (s)

  • 8/8/2019 39248830 Expert Systems

    29/82

    TheThe actionaction (s)(s) areare executedexecuted whenwhen thethe conditioncondition

    (s)(s) ofof thethe rulerule areare satisfiedsatisfied.. TheThe rulesrules createdcreated inin

    thisthis wayway areare collectivelycollectively calledcalled thethe knowledgeknowledgebase/base/ rulerule basedbased systemsystem..

  • 8/8/2019 39248830 Expert Systems

    30/82

    b).b). InferenceInference

    Inference is the drawing of conclusion fromInference is the drawing of conclusion from

    premises.premises.

    Various forms of inferenceVarious forms of inference::

    DeduDeductive and Inductivective and Inductive inference being two of theinference being two of themost commonly used.most commonly used.

  • 8/8/2019 39248830 Expert Systems

    31/82

    Deductive inferenceDeductive inference

    ExampleExample::

    SalmanSalman isis aa programmer programmer ((11))

    AllAll programmersprogrammers areare happyhappy ((22))ThereforeTherefore SulmanSulman isis happyhappy ((33))

    TheThe conclusionconclusion ((33)) inin thisthis exampleexample isis aalogicallogical consequenceconsequence ofof oror cancan bebe deduceddeduced fromfromthethe premises premises ((11)) andand ((22))..

  • 8/8/2019 39248830 Expert Systems

    32/82

    InIn thisthis example,example, thethe conclusionconclusion cancan bebeinferredinferred solelysolely fromfrom thethe informationinformation givengiven.. ThisThis

    deductiondeduction isis aa consequenceconsequence ofof entailmententailment thatthat

    leadsleads toto aa formform ofof reasoningreasoning thatthat isis

    mathematicallymathematically exactexact.. ThisThis meansmeans thatthat ifif thethepremises premises areare truetrue thethe conclusionconclusion isis guaranteedguaranteed

    alsoalso toto bebe truetrue..

  • 8/8/2019 39248830 Expert Systems

    33/82

    Inductive inferenceInductive inference

    ExampleExample::

    AllAll animalsanimals eateat ((44))

    ThereforeTherefore allall kangarooskangaroos eateat ((55))

    OnOn firstfirst inspection,inspection, itit mightmight bebe temptingtempting totoassumeassume thatthat thethe aboveabove conclusionconclusion ((55),), namelynamely allall

    kangarooskangaroos eat,eat, followsfollows by by deductiondeduction fromfrom thethepremisepremise ((44)).. HoweverHowever thisthis assumptionassumption isis falsefalse forforthethe premisepremise givengiven sayssays nothingnothing aboutabout whetherwhether orornotnot aa kangarookangaroo isis anan animalanimal..

  • 8/8/2019 39248830 Expert Systems

    34/82

    Hence,Hence, thethe conclusionconclusion ((55)) doesdoes notnot havehave thethesamesame mathematicalmathematical exactnessexactness thatthat thethe previousprevious

    conclusionconclusion ((33)) hadhad.. ForFor thethe conclusionconclusion ((55)) toto bebe

    truetrue wewe wouldwould havehave toto induceinduce thethe assumptionassumption

    thatthat anan animalanimal isis aa kangarookangaroo toto thethe givengivenpremisepremise ((44)) fromfrom thethe commoncommon--sensesense knowledgeknowledge

    ofof thethe worldworld.. ForFor thisthis reason,reason, suchsuch inferencesinferences areare

    calledcalled inductiveinductive inferencesinferences..

  • 8/8/2019 39248830 Expert Systems

    35/82

    Declarative Knowledge:Declarative Knowledge:

    Example:Example:

    Aslam likes tennisAslam likes tennis (6)(6)

    Everything that Aslam likes, Saleem likesEverything that Aslam likes, Saleem likes (7)(7)

    DeductiveDeductive reasoningreasoning leadsleads toto thethe conclusionconclusion

    thatthat SaleemSaleem likeslikes tennistennis givengiven thethe aboveabove premisespremises..TheThe rulerule ((77)) andand factfact ((66)) thatthat areare givengiven leadlead toto thisthis

    conclusionconclusion..

  • 8/8/2019 39248830 Expert Systems

    36/82

    TheThe conclusionconclusion isis foundfound by by reasoningreasoning withwith

    whatwhat isis given,given, notnot byby explicitlyexplicitly programmingprogramming thethe

    computercomputer onon howhow toto reachreach thethe solutionsolution..

    AnAn exampleexample ofof aa languagelanguage thatthat usesuses thethedeclarativedeclarative paradigm paradigm isis thethe AIAI languagelanguage

    PROLOGPROLOG (programming(programming logic)logic);; itit containscontains anan inin--

    builtbuilt logiclogic toto deducededuce solutionsolution toto thethe problemsproblems ofof

    thethe declarativedeclarative typetype..

  • 8/8/2019 39248830 Expert Systems

    37/82

    Procedure knowledgeProcedure knowledge

    ConventionalConventional algorithmicalgorithmic paradigm paradigm thatthat

    describesdescribes howhow aa procedureprocedure isis executedexecuted bybyexplicitlyexplicitly describingdescribing thethe stepssteps involvedinvolved reachingreaching aa

    solutionsolution..

  • 8/8/2019 39248830 Expert Systems

    38/82

    Inference engineInference engine

    HumanHuman expertexpert whowho havehave toto knowknow notnot onlyonly

    whichwhich rulesrules toto apply,apply, butbut alsoalso inin whatwhat orderorder theythey

    shouldshould be be appliedapplied toto solvesolve aa particular particular problemproblem..SimilarlySimilarly anan expertexpert systemsystem needneed toto decidedecide whichwhich

    rules,rules, andand inin whatwhat order,order, shouldshould be be selectedselected forfor

    evaluationevaluation.. ToTo thisthis anan expertexpert systemsystem usesuses anan

    inferenceinference engineengine..

  • 8/8/2019 39248830 Expert Systems

    39/82

    OrOr inin otherother wordswords

    InferenceInference engineengine isis aa programprogram thatthat interpretsinterprets thethe

    rulesrules inin thethe knowledgeknowledge base base inin orderorder toto drawdraw

    conclusionsconclusions..

    OrOr inin otherother wordswords

    InferenceInference engine,engine, aa componentcomponent ofof expertexpert systemsystem

    controlscontrols howhow andand whenwhen thethe informationinformation inin thetheknowledgeknowledge basebase isis appliedapplied

  • 8/8/2019 39248830 Expert Systems

    40/82

    TwoTwo mainmain alternativealternative strategiesstrategies areare useuse withwith

    ruleruleb

    asedb

    ased systemsystem

    a)a) BackwardBackward--chainingchaining InferenceInference EngineEngine

    ThisThis isis aa goalgoal orientedoriented inin thethe sensesense thatthat itit triestries

    toto proveprove aa goalgoal oror rulerule

    conclusionconclusion byby confirmingconfirming thethe truthtruth ofof allall ofof itsitspremisespremises..

  • 8/8/2019 39248830 Expert Systems

    41/82

    IFIF thethe stainstain ofof thethe organismorganism isis gramgramnegativenegative

    ANDAND thethe morphologymorphology ofof thethe organismorganism isis rodrod

    ANDAND thethe aerobicityaerobicity ofof thethe organismorganism isisanaerobicanaerobic

    THENTHE N therethere isis stronglystrongly suggestivesuggestive evidenceevidencethatthat thethe classclass ofof thethe organismorganism isis

    EnterobacteriaceaeEnterobacteriaceae

    InIn thisthis example,example, thethe inferenceinference engineengine triestries totoproveprove thethe conclusionconclusion ofof thethe rulerule byby attemptingattempting totoproveprove eacheach premisepremise leadingleading toto thethe rulerule conclusionconclusion..

  • 8/8/2019 39248830 Expert Systems

    42/82

    b)b) ForwardForward--ChainingChaining InferenceInference EngineEngine..

    ItIt examinesexamines thethe currentcurrent statestate ofof thethe knowledgeknowledge

    basebase andand finds finds thosethose rulesrules thatthat whosewhose premisespremisescancan bebe satisfiedsatisfied fromfrom knownknown givengiven datadata andand addsadds

    thethe conclusionsconclusions ofof thosethose rulesrules toto thethe knowledgeknowledge

    basebase

  • 8/8/2019 39248830 Expert Systems

    43/82

    forfor examplesexamples ::

    IfIf MalariaMalaria isis diagnosed,diagnosed, teststests etcetc.. maymay bebe

    requestedrequested forfor confirmationconfirmation..

    ThiefThief couldcould be be prosecuted prosecuted inin thethe thefttheft casecase

    throughthrough CourtCourt hearingshearings andand trialstrials etcetc..

    UserUser InterfaceInterface

    TheThe componentcomponent ofof anan expertexpert system system thatthatcommunicatescommunicates withwith thethe useruser isis knownknown asas useruser

    interfaceinterface..

  • 8/8/2019 39248830 Expert Systems

    44/82

    AnAn expertexpert systemsystem useruser interfaceinterface isis normallynormally ofof

    aa highlyhighly interactiveinteractive nature,nature, toto reflectreflect thethe formform ofof

    dialoguedialogue thatthat taketake placeplace betweenbetween aa clientclient andand aa

    humanhuman expertexpert..

    UserUser interfaceinterface willwill notnot onlyonly enableenable thethe useruser totoansweranswer questionsquestions but but alsoalso allowallow thethe useruser toto

    interruptinterrupt itsits runningrunning operationoperation by by askingasking forfor

    explanationsexplanations..

  • 8/8/2019 39248830 Expert Systems

    45/82

    ForFor example,example, aa useruser ofof aa medicalmedical expertexpertsystemsystem mightmight wantwant toto knowknow whywhy systemsystem asksasks forfor

    thethe patientspatients ageage.. ThisThis typetype ofof interrogativeinterrogative isiscalledcalled aa WhyWhy explanationexplanation..

    AnotherAnother typetype ofof explanationexplanation interrogativeinterrogative isis

    thethe HowHow questionquestion.. ThisThis isis normallynormally usedused totoquestionquestion howhow thethe systemsystem arrivedarrived atat aa particularparticulardecisiondecision..

    For For example,example, ifif aa useruser ofof aa medicalmedical expertexpertsystemsystem isis toldtold thatthat thethe patientpatient hashas meningitis,meningitis, thetheuseruser mightmight wantwant toto knowknow howhow thethe systemsystem arrivedarrivedatat thisthis conclusionconclusion..

  • 8/8/2019 39248830 Expert Systems

    46/82

    Architecture OfAn Expert SystemArchitecture OfAn Expert System

  • 8/8/2019 39248830 Expert Systems

    47/82

    DECISIONSUPPORTSYSTEMDECISIONSUPPORTSYSTEM

    DecisionDecision supportsupport systemssystems (DSS)(DSS) areare aa branch branch ofofOperationalOperational ResearchResearch (OR),(OR), whichwhich evolvedevolved fromfrom thetheneedneed toto applyapply quantitativequantitative techniquestechniques toto thethe solutionsolutionofof complexcomplex managementmanagement problemsproblems..

    DSSDSS areare traditionallytraditionally definedefine asas thethe applicationapplication ofof

    OperationalOperational ResearchResearch..

  • 8/8/2019 39248830 Expert Systems

    48/82

  • 8/8/2019 39248830 Expert Systems

    49/82

    DECISION SUPPORT SYSTEM and EXPERTDECISION SUPPORT SYSTEM and EXPERT

    SYSTEMSYSTEM

    DSSDSS supportsupport decisiondecision makersmakers usingusing datadata processingprocessingand/orand/or OROR techniquestechniques;; ExpertExpert systemsystem makemake thethedecisiondecision themselvesthemselves..

    ExpertExpert systemssystems areare notnot appliedapplied toto problemsproblems requiringrequiringthethe mathematicalmathematical optimizationoptimization techniquestechniques thatthat areare

    appliedapplied toto traditionaltraditional DSSDSS..

  • 8/8/2019 39248830 Expert Systems

    50/82

    ExpertExpert systemssystems cancan bebe appliedapplied toto problemsproblems whosewhoseobjectivesobjectives andand constraintsconstraints areare difficultdifficult toto specifyspecify ininquantityquantity termsterms..

    ExpertExpert systemssystems areare effectiveeffective forfor elicitingeliciting alternativesalternatives

    asas part part ofof thethe solutionsolution processprocess;;

    thesethese maymay includeincludewhatwhat ifif scenariosscenarios etcetc.. ThisThis isis duedue toto thethe reasoningreasoningcapabilitiescapabilities ofof expertexpert systemssystems..

    DSSDSS oftenoften aidaid humanhuman decisiondecision makingmaking by by somesomequalitativequalitative oror quantitiesquantities analysis,analysis, whereaswhereas expertexpertsystemsystem trytry toto replacereplace thethe humanhuman expertiseexpertise oror expertexpert..

  • 8/8/2019 39248830 Expert Systems

    51/82

    Decision Support System Verses Expert SystemDecision Support System Verses Expert System

    AttributesAttributes DSSDSS Expert systemExpert system

    ObjectivesObjectives

    Major orientationMajor orientation

    Decision makerDecision maker

    Problem domainProblem domain

    World viewWorld view

    Query directionQuery direction

    Data manipulationData manipulation

    Mathematical modelsMathematical models

    IncorporatedIncorporated

    Assist decision makingAssist decision making

    Decision makingDecision making

    User and/or systemUser and/or system

    Broad, complexBroad, complex

    OpenOpen

    Human queriesHuman queries

    Numerical/symbolicNumerical/symbolic

    --

    Copy and replaceCopy and replace

    Human CapabilitiesHuman Capabilities

    Transfer of experienceTransfer of experience

    SystemSystem

    NarrowNarrow

    ClosedClosedSystem queries humanSystem queries human

    SymbolicSymbolic

    --

  • 8/8/2019 39248830 Expert Systems

    52/82

    EXPERTSYSTEMIN BUSSINESSESEXPERTSYSTEMIN BUSSINESSES

    SupposeSuppose youyou ownown aa realreal estateestate business,business,andand youyou generategenerate overover 4040 percent percent ofof youryourrevenuerevenue fromfrom appraisingappraising commercialcommercial realreal

    estateestate.. ConsiderConsider furtherfurther thatthat onlyonly oneone personperson ininyouryour firmfirm isis capablecapable ofof performing performing thesetheseappraisalsappraisals.. WhatWhat ifif thatthat personperson werewere toto quitquit ??HowHow dodo youyou replacereplace thatthat expertise?expertise? HowHow fastfast

    cancan youyou findfind someonesomeone elseelse ?? HowHow muchmuchbusinessbusiness wouldwould youyou loselose ifif itit tooktook youyou aa monthmonthtoto findfind aa suitablesuitable replacementreplacement ??

  • 8/8/2019 39248830 Expert Systems

    53/82

    InIn business, business, peoplepeople areare valuablevaluable becausebecause theythey

    performperform importantimportant businessbusiness taskstasks.. ManyMany ofof thesethesebusiness business taskstasks requirerequire expertise,expertise, andand peoplepeople oftenoften

    carrycarry expertiseexpertise inin theirtheir headsheads andand thatsthats thethe onlyonly

    placeplace itit cancan bebe foundfound inin thethe organizationorganization.. AIAI cancan

    provideprovide youyou withwith anan expertexpert systemsystem thatthat cancan capturecaptureexpertise,expertise, thusthus makingmaking itit availableavailable toto thosethose whowho areare

    expertsexperts soso thatthat theythey cancan useuse it,it, eithereither toto solvesolve aa

    problemproblem oror toto learnlearn howhow toto solvesolve aa problemproblem..

  • 8/8/2019 39248830 Expert Systems

    54/82

    APPLICATIONSOFEXPERTSYSTEMAPPLICATIONSOFEXPERTSYSTEM

    AnAn expertexpert systemsystem isis usuallyusually builtbuilt forfor aa specificspecific

    applicationapplication areaarea calledcalled aa domaindomain.. YouYou cancan findfind expertexpert

    systemssystems inin thethe followingfollowing domainsdomains::

    AccountingAccounting for for auditing,auditing, taxtax planning,planning,

    managementmanagement consultingconsulting andand TrainingTraining etcetc..

    MedicineMedicine toto prescribe prescribe antibioticsantibiotics wherewhere manymanyconsiderationsconsiderations mustmust bebe takentaken intointo accountaccount (such(such asas

    thethe patientspatients medicalmedical history,history, thethe sourcesource ofof infection,infection,

    thethe mostmost inexpensiveinexpensive drugdrug available)available)

  • 8/8/2019 39248830 Expert Systems

    55/82

    ProcessProcess controlcontrol forfor example,example, toto controlcontrol offsetoffset

    lithographiclithographic printingprinting HumanHuman ResourceResource ManagementManagement toto helphelp personnelpersonnel

    managersmanagers determinedetermine whetherwhether theythey areare inin

    compliancecompliance withwith anan arrayarray ofof federalfederal employmentemployment

    lawslaws..

    FinancialFinancial ManagementManagement -- toto identifyidentify delinquencydelinquency--

    proneprone accountsaccounts inin thethe loanloan departmentsdepartments ofof banksbanks..

    ProductionProduction -- toto guideguide thethe manufacturemanufacture ofof allall sortssorts ofofproducts,products, suchsuch asas aircraftaircraft parts,parts, andand soso onon

    ForestryForestry ManagementManagement -- toto helphelp withwith harvestingharvesting

    timbertimber onon forestforest landslands

  • 8/8/2019 39248830 Expert Systems

    56/82

    Expert System to Control Traffic LightExpert System to Control Traffic Light

    LetsLets looklook atat aa veryvery simplesimple expertexpert systemsystem thatthat

    wouldwould telltell aa driverdriver whatwhat toto dodo whenwhen approachingapproaching aa

    traffictraffic lightlight.. DealingDealing withwith traffictraffic lightlight isis thethe typetype ofofproblemproblem toto whichwhich anan expertexpert systemsystem isis wellwell -- suitedsuited..

    ItIt isis aa recurringrecurring problem,problem, andand toto solvesolve itit youyou followfollow

    aa wellwell--defineddefined setset ofof stepssteps.. YouYou havehave probablyprobably gonegone

    throughthrough thethe followingfollowing mentalmental questionquestion andand answeranswersessionsession many,many, manymany timestimes withoutwithout eveneven realizingrealizing itit..

  • 8/8/2019 39248830 Expert Systems

    57/82

    WhenWhen youyou approachapproach aa greengreen traffictraffic light,light, youyouproceedproceed onon throughthrough.. IfIf thethe lightlight isis red,red, youyou trytry toto stopstop..

    IfIf youyou areare unableunable toto stop,stop, andand ifif traffictraffic isis approachingapproaching

    fromfrom thethe leftleft oror right,right, youyou willwill probablyprobably byby inin troubletrouble..

    Similarly,Similarly, ifif thethe lightlight isis yellow,yellow, youyou maymay bebe ableable totomakemake itit throughthrough thethe intersectionintersection beforebefore thethe lightlight turnsturns

    redred.. IfIf not,not, youyou willwill againagain bebe facedfaced withwith problemproblem ofof

    approachingapproaching traffictraffic..

  • 8/8/2019 39248830 Expert Systems

    58/82

    Expert System to Control Traffic LightExpert System to Control Traffic Light

    RuleSymptomRuleSymptomorFactorFact

    YesYes NoNo ExplanationExplanation

    1). Is the light1). Is the light

    green?green?

    Go through theGo through the

    Intersection.Intersection.

    Go to Rule2.Go to Rule2. ShouldShould be be safesafe ifif

    lightlight isis GreenGreen ..IfIf

    not,not, needneed moremore

    informationinformation..

    2). Is the light2). Is the light

    red?red?

    Go to Rule 4Go to Rule 4 Go to Rule 3Go to Rule 3 Should stop, mayShould stop, may

    not be able to.not be able to.

  • 8/8/2019 39248830 Expert Systems

    59/82

    33)).. IsIs thethe lightlight

    likelylikely toto ChangeChange

    toto redred beforebefore

    youyou getget throughthrough

    thethe Intersection?Intersection?

    Go to Rule 4Go to Rule 4 GoGo throughthrough thethe

    intersectionintersection

    WillWill onlyonly reachreach

    thisthis pointpoint ifif lightlight

    isis yellow,yellow, thenthenyouyou willwill havehave

    twotwo choiceschoices..

    44)).. CanCan youyou stopstop

    before before enteringenteringthethe intersection?intersection?

    StopStop Go to Rule 5Go to Rule 5 ShouldShould stop,stop, butbut

    therethere maymay be be aaproblem problem ifif youyou

    cantcant stopstop..

    55)).. IsIs traffictraffic

    approachingapproachingfromfrom eithereither sideside

    PreparePrepare toto crashcrash GoGo throughthrough thethe

    IntersectionIntersection

    UnlessUnless thethe

    intersectionintersection isisclearclear ofof traffictraffic

    youyou areare likelylikely toto

    crashcrash

  • 8/8/2019 39248830 Expert Systems

    60/82

    WHATEXPERTSYSTEMSCAN DOWHATEXPERTSYSTEMSCAN DO

    AnAn expertexpert systemsystem usesuses ITIT toto capturecapture andand applyapplyhumanhuman expertiseexpertise.. ForFor problemsproblems withwith clearclear rulesrules andand

    procedures,procedures, expertexpert systemssystems workwork veryvery wellwell andand cancanprovide provide youryour companycompany withwith greatgreat advantagesadvantages.. AnAnexpertexpert systemsystem cancan::

    HandleHandle massivemassive amountsamounts ofof informationinformation

    ReduceReduce errorserrors

  • 8/8/2019 39248830 Expert Systems

    61/82

    AggregateAggregate informationinformation fromfrom variousvarious sourcessources

    ImproveImprove customercustomer serviceservice

    ProvideProvide consistencyconsistency inin decisiondecision makingmaking

    ProvideProvide newnew informationinformation

    DecreaseDecrease personnelpersonnel timetime spentspent onon taskstasks

    ReduceReduce costcost

  • 8/8/2019 39248830 Expert Systems

    62/82

    WhereExpert Systems AreMost EffectiveWhereExpert Systems AreMost Effective

    ExsysExsys IncInc.. developsdevelops andand implementsimplements advisoryadvisorysystemsystem whichwhich areare distributeddistributed viavia thethe Internet,Internet,

    IntranetsIntranets oror throughthrough standstand--alonealone technologytechnology.. PotentialPotentialcustomers,customers, supportsupport staff,staff, clientsclients andand employeesemployees areareableable toto interactinteract withwith thesethese systemssystems toto obtainobtainrecommendations,recommendations, asas ifif theythey werewere consultingconsulting withwith aafirms'firms' BestBest ExpertsExperts -- andand they'rethey're availableavailable 2424 hourshours aa

    dayday..

  • 8/8/2019 39248830 Expert Systems

    63/82

    Exsys'sExsys's approachapproach andand techniquestechniques ofofknowledgeknowledge codificationcodification andand representationrepresentation havehave

    beenbeen highlyhighly successfulsuccessful overover thethe lastlast 1717 yearsyears withwith

    thousandsthousands ofof advisoryadvisory systemssystems inin useuse worldwideworldwide..

    AsAs partpart ofof thethe preliminarypreliminary processprocess ofof workingworkingwithwith youryour firm,firm, ExsysExsys firstfirst identifiesidentifies thethe

    appropriateappropriate projectproject areasareas thatthat willwill benefitbenefit mostmost byby

    implementingimplementing advisoryadvisory systemssystems..

  • 8/8/2019 39248830 Expert Systems

    64/82

    ToTo getget thethe bestbest useuse ofof expertexpert knowledgeknowledge inin advisoryadvisorysystems,systems, thethe problemproblem--solvingsolving logiclogic shouldshould::

    .... BeBe wellwell understoodunderstood andand documenteddocumented

    .... BeBe basedbased onon logicallogical steps/procedures/businesssteps/procedures/business rulesrules

    .... Not Not involveinvolve intuition,intuition, guesses,guesses, oror be be based based ononarbitraryarbitrary "personal"personal taste"taste" decisionsdecisions

  • 8/8/2019 39248830 Expert Systems

    65/82

  • 8/8/2019 39248830 Expert Systems

    66/82

    RegulatoryComplianceRegulatoryCompliance

    ThoughThough thethe knowledgeknowledge inin thisthis areaarea cancan bebe veryverycomplex,complex, regulationsregulations areare generallygenerally documenteddocumented inin aaformform veryvery similarsimilar toto thethe IF/THENIF/THEN formform ofof businessbusinessrulesrules.. RegulatoryRegulatory systemssystems insureinsure thatthat allall possiblepossible

    relevantrelevant regulationsregulations areare consideredconsidered andand allall policiespoliciesfollowedfollowed consistentlyconsistently.. TheseThese systemssystems provideprovidesignificantsignificant costcost savingssavings byby helpinghelping companiescompanies staystaywithinwithin industryindustry compliance,compliance, protectprotect employees,employees, avoidavoidpotential potential fines,fines, andand possiblepossible badbad publicitypublicity.. WhetherWhether

    onlineonline oror runrun asas aa download,download, compliancecompliance advisoryadvisorysystemssystems provideprovide aa moremore personalizedpersonalized andand confidentialconfidentialenvironmentenvironment andand interfaceinterface..

  • 8/8/2019 39248830 Expert Systems

    67/82

    Product Selection /Product Selection / RecommendationRecommendation

    SelectingSelecting whichwhich productsproducts bestbest meetmeet aa customerscustomersneedsneeds andand requirementsrequirements cancan be be aa veryvery intricateintricateprocessprocess.. ButBut itit isis oneone thatthat cancan bebe expressedexpressed inin logiclogic

    rulesrules relatingrelating toto customercustomer needsneeds andand productproductspecificationsspecifications.. UnlikeUnlike casecase--based based oror "learning""learning"approaches,approaches, expertexpert systemssystems handlehandle conflictingconflictingrequirementsrequirements andand alwaysalways givegive aa recommendationrecommendation ofof

    thethe bestbest fit,fit, eveneven whenwhen allall customercustomer desiresdesires cancan notnot bebemetmet.. AdvisoryAdvisory systemssystems alsoalso makemake itit possiblepossible forfor staffstafftoto identifyidentify crosscross--sellingselling opportunitiesopportunities andand bebe ableable totosellsell aa muchmuch broader,broader, moremore complexcomplex productproduct lineline..

  • 8/8/2019 39248830 Expert Systems

    68/82

    ConfigurationConfiguration

    ConfiguringConfiguring complexcomplex systemssystems withwith manymany piecespiecesisis veryvery complexcomplex.. DeterminingDetermining whichwhich pieces pieces arearerequiredrequired andand whichwhich areare incompatibleincompatible requiresrequires detaileddetailed

    knowledgeknowledge andand analysisanalysis.. ThisThis isis anan idealideal problemproblem forforanan expertexpert system,system, whichwhich cancan guaranteeguarantee validvalid andandcompletecomplete configurationsconfigurations.. TheThe systemsystem cancan alsoalso explainexplainthethe reasonreason forfor itsits recommendationrecommendation.. DifferentDifferent aspectsaspectscancan comecome intointo playplay atat thethe samesame time,time, forfor instanceinstance::

    inventory,inventory, currentcurrent pricingpricing andand customercustomer requirementsrequirements..TheThe expertexpert systemsystem alsoalso enablesenables comparisoncomparison ofof "what"whatif"if" scenariosscenarios..

  • 8/8/2019 39248830 Expert Systems

    69/82

    ProblemProblem--Solving DiagnosticsSolving Diagnostics

    WhenWhen expertsexperts identifyidentify malfunctionsmalfunctions oror interpretinterpretcomplexcomplex data,data, theythey quicklyquickly looklook forfor symptomssymptoms indicativeindicative ofofparticularparticular problemsproblems.. TheThe knowledgeknowledge ofof howhow toto handlehandle thesetheseproblemsproblems isis idealideal forfor conversionconversion intointo anan advisoryadvisory systemsystem..

    ThereThere areare usuallyusually manymany domaindomain expertsexperts throughoutthroughout aabusinessbusiness enterpriseenterprise.. TheyThey areare oftenoften repeatedlyrepeatedly interruptedinterrupted totoansweranswer commoncommon questionsquestions byby thosethose withwith lessless expertiseexpertise..AdvisoryAdvisory systemssystems freefree upup expertsexperts toto handlehandle moremore complexcomplexproblemsproblems andand projects projects byby makingmaking thisthis knowledgeknowledge accessibleaccessible

    byby employeesemployees oror customerscustomers thatthat needneed itit.. TheseThese systemssystems arearealsoalso beneficialbeneficial inin capturingcapturing andand codifyingcodifying thethe problemproblem --solvingsolving expertiseexpertise ofof thesethese bestbest peoplepeople thatthat maymay bebe retiringretiring ororchangingchanging jobsjobs..

  • 8/8/2019 39248830 Expert Systems

    70/82

    Data AnalysisData Analysis

    ThereThere areare manymany typestypes ofof datadata analysisanalysis whichwhich requirerequire

    bothboth numericnumeric analysisanalysis andand complexcomplex logiclogic analysisanalysis ofof thethe

    datadata.. ExpertExpert systemssystems areare anan idealideal wayway toto dealdeal withwith suchsuch

    problemsproblems.. TheThe purely purely numericnumeric analysisanalysis cancan be be handledhandled

    internally,internally, oror itit cancan bebe interfacedinterfaced toto otherother externalexternal programsprograms..

    TheThe additionaddition ofof powerfulpowerful rulerule--basedbased logiclogic analysisanalysis makesmakes itit

    possible possible toto handlehandle veryvery complexcomplex problems problems thatthat werewere

    previouslypreviously impracticalimpractical.. PersonsPersons interactinginteracting withwith thethe systemssystems

    thenthen don'tdon't requirerequire thethe extensiveextensive knowledgeknowledge ofof interpretinginterpretingdatadata.. TheThe expertexpert systemsystem analysisanalysis process process cancan alsoalso bebe

    completelycompletely embeddedembedded withinwithin otherother systemssystems toto appearappear

    invisibleinvisible toto endend--useruser..

  • 8/8/2019 39248830 Expert Systems

    71/82

    Customer/Product SupportCustomer/Product Support

    TheseThese problemsproblems areare inin mostmost waysways similarsimilar toto diagnosticdiagnosticproblems, problems, but but oftenoften gogo beyond beyond thethe diagnosisdiagnosis ofof aamalfunctionmalfunction andand includeinclude followingfollowing aa precise precise sequencesequence ororspecificspecific policypolicy andand proceduresprocedures.. ManyMany otherother typestypes ofof "help"help

    desk"desk" softwaresoftware givegive aa guessguess atat aa possible possible solutionsolution -- thetheexpertexpert systemsystem approachapproach willwill consistentlyconsistently givegive thethe bestbestrecommendationrecommendation.. .. TheyThey onlyonly askask pertinent pertinent questionsquestionsinsteadinstead ofof forcingforcing aa customercustomer toto gogo throughthrough aa longlong listlist ofofQ&AQ&A toto comecome upup withwith aa solutionsolution.. AdvisoryAdvisory systemssystems withinwithinhelphelp desksdesks helphelp bring bring lessless experiencedexperienced staffstaff upup toto speedspeed

    quicklyquickly withoutwithout repeatedrepeated trainingtraining oror interruptionsinterruptions.. OnlineOnlinecustomercustomer helphelp systemssystems provide provide aa moremore personalizedpersonalizedenvironmentenvironment andand interface,interface, andand theirtheir automationautomation helpshelps freefreeadvisorsadvisors toto provideprovide valuevalue--addedadded "emotional""emotional" supportsupport..

  • 8/8/2019 39248830 Expert Systems

    72/82

    Background MonitoringBackground Monitoring

    InIn additionaddition toto beingbeing interactive,interactive, expertexpert systemssystems logicallogical

    processingprocessing powerpower cancan bebe runrun inin thethe backgroundbackground toto monitormonitor

    datadata streamsstreams.. TheyThey cancan be be constantlyconstantly analyzinganalyzing thethe datadata

    streamsstreams (often(often realreal--time)time) forfor developingdeveloping problems,problems, specialspecial

    opportunities,opportunities, oror otherother informationinformation thatthat shouldshould bebe

    immediatelyimmediately brought brought toto someone'ssomeone's attentionattention.. TheyThey areare

    especiallyespecially usefuluseful forfor identifyingidentifying veryvery uncommonuncommon problemsproblems

    thatthat fewfew knowknow howhow toto solve,solve, andand cancan preventprevent whatwhat couldcould bebe

    catastrophiccatastrophic resultsresults.. TheseThese systemssystems cancan alsoalso alertalert anotheranothersystemsystem toto veryvery rarerare situationssituations thatthat normallynormally wouldwould bebe tootoo

    monotonousmonotonous oror expensiveexpensive toto monitormonitor..

  • 8/8/2019 39248830 Expert Systems

    73/82

    Inconsistency DetectionInconsistency Detection

    ExpertExpert systemssystems cancan checkcheck customercustomer datadata againstagainst

    policiespolicies andand proceduresprocedures toto detectdetect inconsistenciesinconsistencies thatthat

    maymay indicateindicate problems problems likelike fraudfraud oror otherotherirregularitiesirregularities.. TheThe abilityability ofof thethe expertexpert systemssystems toto

    handlehandle complexcomplex logiclogic allowsallows suchsuch systemssystems toto bebe muchmuch

    "smarter","smarter", both both inin detectiondetection andand inin recognitionrecognition ofof

    illegitimateillegitimate usersusers thatthat mightmight triggertrigger otherother systemssystems..

  • 8/8/2019 39248830 Expert Systems

    74/82

    Smart QuestionnairesSmart Questionnaires

    OneOne ofof commoncommon usesuses ofof expertexpert systemssystems isis totomakemake questionnairesquestionnaires moremore intelligentintelligent.. TheThe logiclogic ofof ananexpertexpert systemsystem leadsleads toto askingasking onlyonly pertinentpertinentquestionsquestions thatthat havehave beenbeen determineddetermined toto bebe relevant,relevant,duedue toto answersanswers alreadyalready providedprovided.. AllAll relevantrelevantquestionsquestions areare asked,asked, andand nono irrelevantirrelevant questionsquestions areareaskedasked.. ThisThis producesproduces aa farfar superiorsuperior useruser interfaceinterface.. InIn

    addition,addition, thethe datadata cancan be be analyzedanalyzed asas itit isis bebecollected,collected, andand thethe systemsystem cancan provide provide appropriateappropriatefollowfollow--upup formsforms oror reportsreports onon thethe flyfly..

  • 8/8/2019 39248830 Expert Systems

    75/82

    Process ControlProcess Control

    ExpertExpert systemssystems havehave aa long,long, provenproven historyhistory ofof useuse

    inin controllingcontrolling processes processes toto detectdetect andand correctcorrect

    problemsproblems beforebefore theythey becomebecome seriousserious.. FromFrom DuPontDuPonttoto EastmanEastman Chemical,Chemical, manymany ofof thethe majormajor industrialindustrial

    companiescompanies relyrely onon expertexpert systemssystems.. TheyThey runrun invisiblyinvisibly

    inin thethe background background toto analyzeanalyze variousvarious datadata tagstags

    throughoutthroughout aa processprocess andand alertalert operatorsoperators ofof potentialpotentialproblemsproblems..

  • 8/8/2019 39248830 Expert Systems

    76/82

    Advantages ofExpert SystemsAdvantages ofExpert Systems

    PermanencePermanence ExpertExpert systemssystems dodo notnot forget,forget, butbut

    humanhuman expertsexperts maymay

    ReproducibilityReproducibility ManyMany copiescopies ofof anan expertexpert systemsystem

    cancan bebe made,made, butbut trainingtraining newnew humanhuman expertsexperts isis timetime--consumingconsuming andand expensiveexpensive

    IfIf therethere isis aa mazemaze ofof rulesrules (e(e..gg.. taxtax andand auditing),auditing),

    thenthen thethe expertexpert systemsystem cancan unravelunravel thethe mazemaze

    EfficiencyEfficiency cancan increaseincrease throughputthroughput andand decreasedecreasepersonnelpersonnel costscosts

  • 8/8/2019 39248830 Expert Systems

    77/82

    EfficiencyEfficiency cancan increaseincrease throughputthroughput andand decreasedecrease

    personnelpersonnel costscosts Although expert systems are expensive to build andAlthough expert systems are expensive to build and

    maintain, they are inexpensive to operatemaintain, they are inexpensive to operate

    Development and maintenance costs can be spreadDevelopment and maintenance costs can be spread

    over many usersover many users

    The overall cost can be quite reasonable whenThe overall cost can be quite reasonable when

    compared to expensive and scarce human expertscompared to expensive and scarce human experts

    Cost savings:Cost savings:

    WagesWages-- (elimination of a room full of clerks)(elimination of a room full of clerks)

    Other costsOther costs -- (minimize loan loss)(minimize loan loss)

  • 8/8/2019 39248830 Expert Systems

    78/82

    ConsistencyConsistency WithWith expertexpert systemssystems similarsimilar

    transactionstransactions handledhandled inin thethe samesame wayway.. TheThe systemsystemwillwill makemake comparablecomparable recommendationsrecommendations forfor likelike

    situationssituations..

    HumansHumans areare influencedinfluenced byby

    RegencyRegency effectseffects (most(most recentrecent informationinformation havinghaving aa

    disproportionatedisproportionate impactimpact onon judgment)judgment)

    PrimacyPrimacy effectseffects (early(early informationinformation dominatesdominates thethe

    judgment)judgment)..

  • 8/8/2019 39248830 Expert Systems

    79/82

    DocumentationDocumentation AnAn expertexpert systemsystem cancan provideprovide

    permanentpermanent documentationdocumentation ofof thethe decisiondecision processprocess CompletenessCompleteness AnAn expertexpert systemsystem cancan reviewreview allall thethe

    transactions,transactions, aa humanhuman expertexpert cancan onlyonly reviewreview aa

    samplesample

    TimelinessTimeliness FraudFraud and/and/ oror errorserrors cancan bebe preventedprevented..

    InformationInformation isis availableavailable soonersooner forfor decisiondecision makingmaking

    BreadthBreadth -- TheThe knowledgeknowledge ofof multiplemultiple humanhuman expertexpert

    cancan bebe combinedcombined toto givegive aa systemsystem moremore breadthbreadth thatthat aasinglesingle personperson isis likelylikely toto achieveachieve

  • 8/8/2019 39248830 Expert Systems

    80/82

    Reduce risk of doing businessReduce risk of doing business

    Consistency of decision makingConsistency of decision makingDocumentationDocumentation

    Achieve ExpertiseAchieve Expertise

    EntryEntry barriersbarriers ExpertExpert systemssystems cancan helphelp aa firmfirm

    createcreate entryentry barriersbarriers forfor potentialpotential competitorscompetitors

    DifferentiationDifferentiation InIn somesome cases,cases, anan expertexpert systemsystem

    cancan differentiatedifferentiate aa productproduct oror cancan bebe relatedrelated toto thethe

    focusfocus ofof thethe firmfirm

    ComputerComputer programsprograms areare bestbest inin thosethose situationssituations wherewhere

    therethere isis aa structurestructure thatthat isis notednoted asas previouslypreviously

    existingexisting oror cancan bebe elicitedelicited

  • 8/8/2019 39248830 Expert Systems

    81/82

    Disadvantages ofExpert SystemDisadvantages ofExpert System

    CommonCommon sensesense InIn additionaddition toto aa greatgreat dealdeal ofof

    technicaltechnical knowledge,knowledge, humanhuman expertsexperts havehave

    commoncommon sensesense.. ItIt isis notnot yetyet knownknown howhow toto givegive

    expertexpert systemssystems commoncommon sensesense..

    CreativityCreativity --HumanHuman expertsexperts cancan respondrespond creativelycreatively toto

    unusualunusual situations,situations, expertexpert systemssystems cannotcannot..

  • 8/8/2019 39248830 Expert Systems

    82/82

    LearningLearning --Human experts automatically adapt toHuman experts automatically adapt tochanging environments, expert systems must be explicitlychanging environments, expert systems must be explicitly

    updated. Caseupdated. Case--based reasoning and neural networks arebased reasoning and neural networks aremethods that can incorporate learning.methods that can incorporate learning.

    SensoryExperienceSensoryExperience-- Human experts have available toHuman experts have available to

    them a wide range of sensory experience; expert systemsthem a wide range of sensory experience; expert systemsare currently dependent on symbolic input.are currently dependent on symbolic input.

    DegradationDegradationExpert systems are not good atExpert systems are not good at

    recognizing when no answer exists or when the problemrecognizing when no answer exists or when the problemis outside their area of expertise.is outside their area of expertise.


Recommended