版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
EstimatingML-modelsFinancialImpact第1题CompletethestatementbelowAprojectmaybeacceptedif…ANPV>0,IRR>costofcapital,PI>1BNPV>=0,IRR>0,PI>0CNPV>=0,IRR>costofcapital,PI>0DNPV=0,IRR>costofcapital,PI>1第2题Tolaunchanewproject,acompanyinvestedUSD30mln.ofequityandtookoutaloanofUSD70mln.for3yearsata5%rate.Costofequityis10%andthetaxrateis20%.CalculatetheWACCvalueInputformat(in%):10.1____第3题Whichofthefollowingisnotavaluationprinciple?AConservatismBSignificanceCMeasurabilityDMateriality第4题Selectthemostimportantcriterioninproject/AIvaluation?AIRRBPPCPIDNPV第5题Youhave2projectswiththefollowingcashflowforecasts.CalculateNPVfortheprojectwiththehighestNPV.Assumethatthediscountrateis10%.YearProjectAProjectB0-300-30011502002100140315050Inputformat:10.1____第6题Youhave2projectswiththefollowingcashflowforecasts.WhatistheIRRfortheprojectwiththelowestNPV?Assumethatthediscountrateis10%.YearProjectAProjectB0-300-30011502002100140315050Inputformat(in%):10.1____第7题Youhave2projectswiththefollowingcashflowforecasts.WhatisthePIfortheprojectwiththehighestNPV?Assumethatthediscountrateis10%.YearProjectAProjectB0-300-30011502002100140315050Inputformat:3.25____第8题IfweknowthatsomeofthecostsandincomeswillnotsignificantlyaffecttheNPV,wedonotneedtowasteourtimetryingtomaketheirassessmentmoreaccurate.Whichprincipleisthis?AMaterialityBConservatismCObjectivityDSensitivity第9题WhatistheaveragevaluationperiodforAIinitiatives?A3yearsstartingfromtheAIsolutiondevelopmentlaunchB3yearsbutitmaytakemoreorlesstimedependingonAImodelfeaturesC5yearsstartingfromtheAIsolutiondevelopmentlaunchD5yearsbutitmaytakemoreorlesstimedependingonAImodelfeatures第10题Whichofthefollowingisnotoneofthe6basicsourcesofincome?AOperationalcostscuttingBWACCreductionCCustomersatisfactiongrowthDLiquiditymanagement第11题Whichstatementiswrong?AWeneedFCFEtousetocalculatePPandDPPBFCFEismoreessentialforbanks,companieswithprimaryequityfinancingCWithoutimplementationAIcodehasnovalue,sovaluationmustinvolvetheassessmentofchangesandcostsrequiredtodevelopandmaintainthesolutionDCannibalizationisalossinotherproductssalescausedbytheboostinsalesforthecertainproduct第12题Wehavethefollowingdataabouttheratesandcompany'smarketsensitivity:risk-freerateis2%,betais1.2,countryriskpremiumis1%,marketriskpremiumis3%.Calculatethecostofequity.Inputformat(in%):10.1____第13题WhichisessentialforcalculatingtheNPV?Pleaseselectallcorrectanswers.AFindingthediscountratedependingonthecashflowBBuildingappropriatecashflowCMakingaconservativeforecastregardingthecosts,risksandestimatedincome第14题Whichstatementiswrong?AIfthePPforprojectA<thePPforprojectB,thentheNPVforprojectAishigherBIfthePPforprojectA<thePPforprojectB,thentheNPVforprojectBishigherCIftheIRRforprojectA<theIRRforprojectB,thentheNPVforprojectAishigherDIftheIRRforprojectA<theIRRforprojectB,thentheNPVforprojectBishigherEAllabove第15题Whichofthefollowingstatementsiswrong?AIfNPVispositivethenIRRishigherthancostofequityBTheIRRismorepreferableapproachofdecisionmakingforprojectswithahighportionofexternalfinancingCBetashowsthecorrelationbetweenthecompany’sstockpricechangeandmarketchangesDCostofequityisnormallyamoreexpensivesourceoffinancingcomparingtothecostofdebt第1题Supposethetotalnumberofcreditworthyclientsis1,000.Theloanmarginis200andtheamountoflossesincaseofdefaultis600.For100clientsthecreditworthinessestimationwaswrong.Claculatethemaximumnumberoffalsepositiveclientswhenyoucanstillavoidnegativeprofits?Inputformat:integer(e.g.250)____第2题ChoosethecorrectstatementAFPandFNcanbereducedsimultaneouslybyadjustingthresholdlevelBFPandFNarepositivelycorrelatedprovidedthemodelisunchangedCFPandFNcanbereducedsimultaneouslybyimprovingthemodelDFPandFNhavezerocorrelationwhentheircostsareequal第3题ChoosethecorrectstatementABenefitcurvecantakeonlynon-negativevaluesBBenefitcurveforarandommodelcannottakevaluesabovezeroCBenefitcurveendpointsalwayshavezerovaluesDItispossiblethatregardlessofmodelqualitythefinancialbenefitisalwaysnotlessthanzero第4题CompletethestatementbelowInordertooptimizecross-sellprofitswithresponsemodeloneshould…ASortallclientsdecreasingbyresponseprobabilitywithcommunicationBSortallclientsascendingbyresponseprobabilitywithcommunicationCSortallclientsdecreasingbydifferencebetweenresponseprobabilitywithandwithoutcommunicationDSortallclientsascendingbydifferencebetweenresponseprobabilitywithandwithoutcommunication第5题Consideracreditscoringproblemwherethepositivecaseisadelinquentclient.FPandFNerrorcostsareequal,numberofpositivecasesis1,000,numberofnegativeonesis500.Supposethatwewillbeacceptingallloanapplications.Whichstatementistrueforthevalueoffinancialbenefit?AFinancialbenefitwillbebelowzeroBFinancialbenefitwillbeabovezeroCFinancialbenefitwillbeequaltozero第6题Considerabinaryclassificationresponsemodelwhichisdeployedtomaximizeprofitswhenofferingaproducttoclients.Suppose,thecallcentercanmakeonly1,000clientoffersperweek.Whichdecisionstrategyseemstobethebesthere?AEstablishcommunicationwithtop1,000clientsperweekwhicharepredictedbymodelasmostprobabletoresponseBEstablishcommunicationwithtop1,000clientsperweekwhicharepredictedbymodelasleastprobabletoresponseCEstablishcommunicationwithcertainnumberoftopclientsperweekwhicharepredictedtohavemaximumincrementinresponseduetocommunication,andtheirexpectedbenefitexceedscommunicationcostsDEstablishcommunicationwithtop1,000clientsperweekwhicharepredictedtohavemaximumincrementinresponseduetocommunication第7题Supposethetotalnumberofcreditworthyclientsis1,000.Loanmarginis100andlossamountincaseofdefaultis2,000.YourcurrentmodelshowsFP=500andFN=10.YourcolleagueofferstousedifferentmodelwhichshowsFP=100andFN=30.Whosemodelisbetterintermsoffinancialbenefit?AThecurrentmodelisbetterBYourcolleague'smodelisbetter第8题Consideracreditscoringmodelwhichaimstopredictloandefaultwhichisdenotedaspositivecase.YourmanagerisworriedaboutdisappointedclientswhowererejectedbasedonmodelpredictionsandsetsagoalforyourdepartmenttoholdFP=0.Whatisgoingtohappenifthisgoalisaccomplished?AThenumberofdefaultsisgoingtodecreaseBThenumberofdefaultsisgoingtoincreaseCTPwillbealsozeroDFNwillbealsozero第9题Supposeyouareeagertoincreaseyourloanmarketsharebyprovidingloanstoasmanyborrowersaspossible.Forsimplicity,leteachloanhaveafixedmargin100andlossgivendefault1000.Youaskyourcreditriskdepartmenttoaccepteveryloanapplication.Analystswarnyouthattheoddsofcreditworthyanduncreditworthyclientsareapproximately7:1.AItistoorisky,youwillhavenegativeprofitsBNevermind,youarestillgoingtobeprofitableandconquermarket第10题SupposeeachclienthasdifferentindividualFPandFNerrorcosts.Youareapplyingacreditscoringmodelthathasdefaultprobabilityasanoutput.Whichofthefollowingstatementsistrue?AYouhavetouseincrementdecisionschemebecauseourdecisionhasimpactonFPandFNofclientsBYouhavetousesimplethresholddecisionstrategybycalculatingoptimalthresholdprobabilityforallclientsCYouhavetobreakclientsingroupswithsimilarFPandFNerrorcostsvaluesDYouhavetousefunctionaldecisionschemeandcalculateoptimalprobabilitythresholdforeachclientQuiz3第1题Whatisamodelrisk?ADeteriorationofmodelqualityonthetestsampleincomparisonwiththetrainingBRiskofadverseconsequencesduetotheuseofmodelsinbusinessprocessesCTheriskthattheclientwillnotreturntheloanDAveragemodelerroronthetrainingsample第2题Thelogisticregressionmodelpredictsthelikelihoodofclientsgoingintodefaultandisusedintheloanapprovalprocess.Whichisnotacaseofmodelriskforsuchamodel?AUnexpectedlyhighFNerrorafterthemodeldeploymentBThequalityofanXGBoostmodelissignificantlyhigherthanthatoflogisticregression.CWemadeamistakewhileimplementingthemodel.Weusedittoperformpredictionsforadifferentclientsegmentcoparingtotrainingdata,whichledtoasharpdecreaseinthenumberofapprovedloansDModelhasshownFNerrorrateclosetotheoneattrainingstage第3题Theestimatedfinancialeffectwhenusingthecurrentmodelis100𝑝𝑒𝑟𝑦𝑒𝑎𝑟.𝐻𝑜𝑤𝑒𝑣𝑒𝑟,𝑡ℎ𝑒𝑟𝑒𝑖𝑠𝑎𝑛𝑎𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑚𝑜𝑑𝑒𝑙𝑜𝑓ℎ𝑖𝑔ℎ𝑒𝑟𝑞𝑢𝑎𝑙𝑖𝑡𝑦.𝑇ℎ𝑒𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙𝑒𝑓𝑓𝑒𝑐𝑡𝑓𝑟𝑜𝑚𝑖𝑡𝑠𝑢𝑠𝑒𝑖𝑠120peryearbutthismodelisnotdeployedduetosomereason.Whatisvalueofmodelriskinthiscase?Inputformat:integer(e.g.250)____第4题Howwouldyouevaluatethemodelriskduetomodelqualitydeteriorationformodelsthatareupdatedonlineusingthemostrecentdata?AZeroBAbovezero第5题Inordertoestimatetheexpectedandunexpecteddeteriorationofmodelquality,weusebootstraptobuildthemodelqualitydistribution.Theinitialsamplesize=N.Howaresamplesextractedduringbootstrap?ASamplesofthesamesizeastheoriginalareextractedwithreplacementB100samplesofsizeN/100C2samplesofsizeN/2DSamplesthatcontainN-1elementsfromtheoriginal第6题Thereisamodelqualitydistribution.Whichquantileofthisdistributionshowsusunexpectedlybadoutcome?Andwhichmetricsdoweuseforexpectedoutcome?AUnexpectedlybad⎯1%upperquantile,expected⎯meanBUnexpectedlybad⎯1%upperquantile,expected⎯1%lowerquantileCUnexpectedlybad⎯1%lowerquantile,expected⎯meanDUnexpectedlybad⎯1%lowerquantile,expected⎯1%upperquantile第7题Whatresearchneedstobedonetocalculatethemodelriskcomponentconnectedwithincorrectuseofthemodel?AWaituntilsufficientstatisticsonerroneousmodeluseisaccumulatedandestimatetheactualdamageBApplyanerroneousandcorrectedmodeltohistoricaldataandcomparethefinancialeffectestimates第8题Whiledevelopingtheresponsemodel,out-of-timesamplingwasused.ThetrainingsamplecontainsinformationaboutcommunicationsfromJanuary2019toDecember2019.Theout-of-timesamplecontainsinformationaboutcommunicationswithinJanuary2020⎯March2020andmodelqualitymetricsis3%lower.PerformthesimplestextrapolationtocalculatetheexpectedmodelqualitydecreasebyDecember2020Inputformat(in%):integer(e.g.25)____第9题Supposethatwhenmodelqualitydecreasesby1%,financialbenefitdecreasesby$10.Theexpecteddecreaseinmodelqualityis2%permonth.Whatisthesimplestestimateofmodelriskcomponentconnectedwithmodelqualitydeteriorationwithinayear?Inputformat(in$):integer(e.g.250)____第10题Suppose,amodelhasbeendevelopedanddeployedinproductionenvironment.Themodelisplannedtobeusedfor5yearswithoutupdating.Totalexpectedbenefitfromitsimplementationduring5yearsis1000.𝐻𝑜𝑤𝑒𝑣𝑒𝑟,𝑡ℎ𝑒𝑟𝑒𝑖𝑠𝑎𝑛𝑒𝑣𝑖𝑑𝑒𝑛𝑐𝑒𝑡ℎ𝑎𝑡𝑡ℎ𝑒𝑚𝑜𝑑𝑒𝑙𝑞𝑢𝑎𝑙𝑖𝑡𝑦𝑖𝑠𝑔𝑜𝑖𝑛𝑔𝑡𝑜𝑑𝑒𝑐𝑟𝑒𝑎𝑠𝑒𝑏𝑦520.Whatwillbenetexpectedbenefit,takingintoaccountthemodelrisk?Inputformat(in$):integer(e.g.250)____Quiz4第1题WhatisthemaingoalofA/Btesting?AToevaluatetheeffectofproposedimprovementinmodel,decisionrules,businessprocessorproductandtoclearitfromothermacroenvironmentalchangesBTolaunchnewdecisionrules,businessprocessesorproductsCTosearchforsimilarclientswithinclient’sbase第2题WhichofthefollowingstatementscontradictswithprinciplesofA/Btesting?AThedifferencebetweenAandBgroupsintermsofexaminedtargeteffectmustbestatisticallysignificantBTheprocessthatistestedviaA/BtestingmustreplicatetheprocessafterscalingthesolutionCThestructureandfeaturesofunitswithinanA/Btestshouldbesignificantlydifferentcomparingtoproductionenvironment第3题WhyisitsoimportanttoperformproperrandomizationforcontrolandtreatmentgroupsbeforeanA/Btest?AInordertokeep“otherthingsequal”principleBInordernottotaketoomuchriskconnectedwiththenewprocessCWeneedrandomizationonlyinasituationwherethereisnoopportunitytoperformanA/Btest第4题WhenpreparingforanA/Btestwemustspecifythemeasurabletarget.Whichofthefollowingishardlymeasurable?AAveragetime-to-marketBSalesCAverageresponserateDClient’smooduplift第5题SupposeweperformanA/Btestforourcorporateclients.Whichfactorscanbeimportanttotakeintoaccountwhenpreparingcontrolandtreatmentgroups?ACompanysizeBBusinessfieldCRegionallocationDAlltheabove第6题WhatshouldwecalculateinordertocheckforproperrandomizationbeforeanA/Btest?AMeans,mediansandvariancesforkeyfeatureswithingroupsAandBBAveragetargetlevelswithingroupsAandBCNumberofobservationswithingroupsAandB第7题Youdevelopedanewmodelthatestimatestheprobabilityofresponse.YoulaunchedanA/Btesttoverifywhetherornotyournewmodelisbetterthanthepreviousone.Youcollectedtheresponses.Whatshouldyoudonow?ACalculatetheaverageresponseratewithinclientswhichwereestimatedwiththenewmodelandrunt-testinordertocheckwhetheritissignificantlydifferentfromzeroBCalculateaveragepredictedresponsesestimatedbynewandoldmodelandrunsamplemeanstestwithnullhypothesisthataveragepredictedresponsesareequalCCalculateaverageactualresponsesseparatelyforclientsestimatedbynewandoldmodelandrunsamplemeanstestwithnullhypothesisthataverageresponsesareequal第8题Whatiscannibalization?ASituationwhenourinitiativestoboostsalesofoneproductorserviceworsenthesalesofourothercompetitiveproductsandservicesBSituationwhenourinitiativestoboostsaleshaveamaximumshort-termeffectatthebeginning(reassessmentofcustomerbase)andthendecreaseCSituationwhenclient’sbehaviormaychangebecauseoftransferringthecostofrisktotheotherparty第9题WhatdiffersWelch’st-testfromStudent’st-test?ATheformerconsiderssamplevariancestobeequalBTheformerisdesignedformultiplesamplesCTheformerconsiderssamplevariancetobesignificantlydifferentDTheformerchecksforvariancesequality,thelatterisusedtocheckformeansequality第10题WhatdiffersBarlett’stestfromF-test?ATheformerconsiderssamplevariancestobeequalBTheybothcheckforvarianceequality,buttheformerisdesignedformultiplesamplesCTheformerconsiderssamplevariancetobesignificantlydifferentDTheformerchecksforvariancesequality,thelatterisusedtocheckformeansequalityQuiz5第1题Supposeyouareimplementinganewmodelwithinthecreditscoringprocess.Whichsourcesofbiascantakeplacewhencollectinghistoricaldatatotrainthemodel?AHistoricaldataisalwaysrepresentative.BThedatayoupossesscontainsinformationonlyaboutclientswhoweregrantedaloanbasedonapreviouslyusedmodel.第2题Whichattributeofrejectedclientsincreditscoring(ortheclientswithnocommunicationinamarketingcampaign)istypicallyunobservable?AInitialfeaturesusedtobuildmodelpredictionsBTargetvariable(defaultorresponseevent)riable(defaultorresponseevent)CModelpredictioncannotbecalculatedforsuchclients第3题HowcanA/Btestinghandletheproblemofbiasedhistoricaldata?ARandomizedcontrolgroupinA/Btestingallowsonetoobservebehaviorof,otherwise,rejectedclientsBA/BtestingimpliesmeasurabletargetCEvaluatingA/Btestresultsinvolveshypothesisstatisticaltestingwhicheliminatesthebias第4题Whenisitimpossibletorunacontrolgroup?AErrorcostsaretooexpensiveBLegislationdoesnotallowtodiscriminateclientsCT-infrastructureispoorDlltheabove第5题WhyisitimportanttoassignweighttoobservationswhenrunningacontrolgroupandrestoringFPandFNerrorratesforthewholeclientbase?ABecausewearenotsureaboutmeasurementaccuracyforsomeobservationsBBecauseusuallyacontrolgroupisarelativelysmallportionofclientsCBecausewewanttofocusonobservationswithhighermodelerrormore第6题Whatcanwesayaboutrejectinferenceasamethod?AItismathematicallyproventhatrejectinferenceallowstofullyrestoreinformationaboutrejectedclientsBUsingthismethodcanhelpustoextractsomeinformationaboutrejectedclien
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 2026高中生物教资面试结构化问答题库及解析
- 上证所固定收益证券综合电子平台交易员培训
- 生产计划与库存管理
- 臀部脓肿健康护理
- 医学顾问职业进阶指南
- 重力式挡土墙施工工艺
- 汽车涂装车间生产安全指南
- 人防外墙穿墙套管防渗注浆指南(2025版)
- 《青鸟》测试题及答案
- 凶残理科试题及答案
- 2025浙江省考评员答题题库(附答案)
- 2025~2026学年七年级上册华东师大版 数学期中测试试卷【含答案】
- 胃癌科普课件
- 专业电子焊接培训课件
- 《深度学习原理及应用》课件 第1章 感知机
- Q-SY 02660-2024 煤层气水平井远距离穿针钻井作业规范
- 死亡游戏主题班会课件
- 2025年度建筑施工安全生产费用提取及使用计划
- 全套电子课件:管理会计(第三版)
- KTV保安服务合同
- 背光异物改善8D
评论
0/150
提交评论