會(huì)計(jì)數(shù)據(jù)分析 Solutions-Manual Chapter-3-EOC-SM-Updated

SolutionsManual–Chapter3

SolutionstoDiscussionQuestions

Whatisthedifferencebetweenatargetandaclass?

Atargetisaspecificattributeorvaluethatananalystistryingtoevaluate,suchasaninterestrateorscore.Aclassisacategoryorgroupingthatadataobjectisassignedto,suchasfraudornotfraud.

Whatisthedifferencebetweenasupervisedandanunsupervisedapproach?

Thesupervisedapproachreliesonananalysisofpastdatatopredicttheclassassignmentorregressedvalueforanewunknownobservation.Classificationandregressionarepopularsupervisedmodels.Anunsupervisedapproachisusedtoexploredataanddiscoverpreviously-unknownpatterns.Clusteringandprofilingarecommonunsupervisedmodelsthathelpresearchersidentifygroupsofdatathatmaynotbeobvious.

Whatisthedifferencebetweentrainingdatasetsandtest(ortesting)datasets?

Supervisedmodelsrelyonpreviously-analyzedhistoricaldatatopredictfutureoutcomes.Forexample,anauditormayidentifyfraudulenttransactionsandlabelthoseasfraud.Aportionofthatdataisusedtotrainthemodel,meaningthatatoolanalyzesthehistoricaltrainingdataandtriestoidentifytheattributesthatarethebestpredictorsofaclassorvalue.Oncethemodelhasbeendeveloped,anotherportionofthehistoricaldataisusedtotestthemodeltoseewhichvaluethemodelpredictsforthatdata.Thetoolthencomparesthepredictedvaluesinthetestdatasetstotheactualvaluesinthetestdatasettoevaluatethemodelforaccuracy.Asetofhistoricaldatacanbesplitmanywaysintotrainingandtestingdatasets.

UsingFigure3-5asaguide,whatarethreedataapproachesassociatedwiththesupervisedapproach?

Classification,Causalmodeling,andregression.

UsingFigure3-5asaguide,whatarethreedataapproachesassociatedwiththeunsupervisedapproach?

Profiling,co-occurrencegrouping,andclustering.

Howmightthedatareductionapproachbeusedinauditing?

Onceanauditorhasidentifiedtypesofdatathatarehighrisk(e.g.transactionsonweekends,vendorswithP.O.Boxaddresses)theymayfilterthedatatoshowonlythosetypesoftransactions(basedonthedate,oraddressfieldinthiscase).

Alsomentionedinthechapterarefilteringonsuspiciousvendornames,sequencechecks,andgapdetection.

Howmightclassificationbeusedinapprovingordenyingapotentialfraudulentcreditcardtransaction?

Inthisanalysis,theclassassignedtoaspecificcreditcardtransactionwouldbeeither“fraud”or“notfraud”.Historicalrecordswouldbeassignedoneofthesetwoclasses,basedoncustomerclaims,etc.Aclassificationmodelwouldusepartofthishistoricaldatatotrainamodeltoidentifytheattributesthatarethebestpredictersofafraudulenttransaction.Thentheremainingdatawouldbeusedtovalidatethemodelandtestforaccuracy.

Howissimilaritymatchingdifferentfromclustering?

Similaritymatchinghasaspecificgoalinmind,suchastryingtofindcustomerswhoarelikeyourbestcustomers.Inthiscase,wehaveaspecifictargetandaretryingtolocatesimilarobjects.Clusteringisanattempttofindnaturalgroupingswithoutbeingdrivenbyaspecificpurpose.Clusteringismoreexploratorywheresimilaritymatchingassumesyouknowwhatyou’relookingfor.

Howdoesfuzzymatchwork?Giveanaccountingsituationwhereitmightbemostuseful?

Afuzzymatchusesprobabilitytoshowlikelymatches,basedonhowmuchthetwovalueshaveincommon.Forexample,tworecordsthatcontainaddresseswithsomedefinedpercentageofmatchingcharacterswouldbeconsideredafuzzymatch.Thisallowsauditorstofindrecordsthatapproximateeachotherinthecasewhereanemployeemighttrytoconcealaconnectionbyvaryingthevaluestoavoidexactmatches.

Compareandcontrasttheprofilingdataapproachandthedevelopmentofstandardcostforaunitofproductionatamanufacturingcompany?Aretheysubstantiallythesameordotheyhavedifferences?

Dataprofilingmaybeusedtodetermineproductioncostandvolumebehaviortodetermineabenchmarkforfuturecostandvolume.Thisislikewhatamanagerofamanufacturingcompanydoesindeterminingstandardcostforaunitofproduction.Theyareverysimilarinthatthegoalistocalculateabenchmarkforcontrollingpurposes.

Themaindifferencesisthatdataprofilingcanincorporatealargeramountofdata(suchasmarkettrends,changingfuelprices,orweatherpatterns)toautomaticallygenerateandcontinuallyupdateamoreprecisebenchmark.

Figures3-1through3-4suggestthatvolumeanddistancearethebestpredictorsof“daystoship”forawholesalecompany?Anyothervariablesthatwouldalsobeusefulinpredictingthenumberof“daystoship”?

Answersvary,butsomesuggestedvariablesmightbenumberofemployeesworking,dayoftheweek,logisticscapacity,temperature,etc.

SolutionstoProblems

Relatedpartytransactionsinvolvepeoplewhohaveclosetiestoanorganization,suchasboardmembers.Assumeanaccountingmanagerdecidesthatfuzzymatchingwouldbeausefultechniquetofindundisclosedrelatedpartytransactions.Whatdatawouldthemanagerneedtotestforrelatedpartytransactions?Whatwouldtheprocesslooklike?

Toperformfuzzymatching,themanagerwouldneedalistofrelatedpartiesandtheircontactinformation.Additionally,shewouldneedthecontactinformationforvendorsandcustomersthatparticipateincompanytransactions.

Themanagerwouldjointherelatedpartycontacttablewiththevendorand/orcustomercontactinformation.Sinceitislikelythattheaddresseswillbesimilarbutnotexact,usingthefuzzymatchtoolinExcelorIDEAwouldhavethemanagerselectthesimilarfields,inthiscaseaddressandzipcode.Themanagerwouldthenreviewthetransactionsthatinvolvevendorsorcustomersthatmatchtoseeiftheyarerelatedpartytransactions.

Anauditoristryingtofigureoutiftheinventoryatanelectronicsstorechainisobsolete.Whatcharacteristicsmightbeusedtohelpestablishamodelpredictinginventoryobsolescence?

Answersmayvary.Theauditormaylookatsimplemetricssuchastheageoftheinventory(e.gbasedonpurchasedate),orratios(e.g.turnoverforspecificproducts).Ifthereisarecordofinventorythathasbeendeemedobsoleteinthepast,theauditorsmaybeabletodevelopamodelbasedoncharacteristicsofthoseitems(e.g.size,type,manufacturer).Aclassificationmodelwoulddeterminetheprobabilityofwhichitemsareobsoleteornotobsoleteandcouldbeusedtoevaluateaclient’scompleteinventory.

Anauditoristryingtofigureoutifthegoodwillitsclientrecognizedwhenitpurchasedafactoryhasbecomeimpaired.Whatcharacteristicsmightbeusedtohelpestablishamodelpredictinggoodwillimpairment?

Goodwillimpairmentiscalculatedusingatwo-steptest.Firsttheauditormustdeterminewhetherthegoodwillisimpairedbycomparingthebookvaluewiththefairvalue.Thentheauditormustcalculatetheimpliedfairvalueofgoodwillandcollectevidenceastowhethermanagementrecordedtheimpairment.

Amodelwouldneedtolookatbothquestionsbasedoninput(e.g.accountbalances)fromthegeneralledgeranddeterminantsoffairvalue(e.g.marketdata,assessmentdata).Tocreateatrulypredictivemodel,theauditorwouldcollectdataonimpairmentfromotherclientsandusethoseobservationstobuildamodelthatcouldbeusedtopredictwhetheranewclientisalsoimpaired.

Thisprovidesaninterestingdiscussiononprivacyconcerns.Forexample,wouldaclientbewillingtosharedatathatcouldbeusedtobuildamodelfortheauditors?Mostlikely,no.Couldtheauditorbuildamodeliftheirclienthadmultipleacquireddivisionswithahistoryofimpairment?Probably,yes,buttheremaynotbesufficientobservationstomakeanaccurateenoughprediction.

Howmightclusteringbeusedtoexplaincustomersthatoweusmoney(accountsreceivable)?

Oneformofclusteringthatisalreadyusedforaccountsreceivableistheagingofaccounts.Theaginggroupsaccountsbyhowoldthereceivableis,withtheexpectationthatolderaccountsarelesslikelytobecollected.

Agingreliesononlyonedimension,time,andfocusesonthetransaction,notthecustomer.Clusteringmaybeusefulindeterminingwhethercustomersformnaturalgroupingsrelativetotheirabilitytopaytheirbills,basedoncorrelatedattributes,suchaslocation,size,volumeoforders.

Ifwehavegooddatathatshowswhichcustomershavehadaccountswrittenoff,wemayexpandthismodeltopredictthelikelihoodofnonpaymentbyusingaclassificationmodel.

Whywouldtheuseofdatareductionbeusefultohighlightrelatedpartytransactions(e.g.,CEOhasherownseparatecompanythatthemaincompanydoesbusinesswith)?

Answerswillvary.Datareductioncanbeusedtofiltertransactionsonspecificattributes.Byremovingunrelatedtransactionsfromtheanalysis,managementoranauditorcouldclearlyseethescopeandvolumeoftransactionsandeitheracceptthosewithadisclosureormakearecommendationtoimplementbetterinternalcontrolstopreventthemfromoccurring.

HowcouldXBRLbeusedbyaninvestortodoananalysisoftheindustry’sinventoryturnover?

AssumingXBRLdataisvalidandaccurate,aninvestorwouldidentifyspecificaccounttags(e.g.InventoryNet,