研发项目风险概率评估方法研究-基于动态贝叶斯网络研究视角-张俊光

2014－02－0671072123201311151972－、1988－。———100083。LeakyNoisy－orgate。。DOI10．13956/j．ss．1001－8409．2015．02．28F224A1001－8409201502－0131－05ＲesearchandDevelopmentProjectＲiskProbabilityAssessment———BasedonDynamicBayesianNetworkZHANGJun-guangXUZhen-chaoSchoolofEconomicsandManagementUniversityofScienceandTechnologyBeijingBeijing100083AbstractAresearchanddevelopmentＲ＆DprojectriskprobabilityassessmentmodelbasedondynamicBayesiannet-workisproposedwhichcombinestimeseriestheoryandBayesiantheory．Throughprobabilityanddirectedacyclicgraphitexpressesthechangingrelationshipofriskfactorsstatebetweendifferenttimesegments．MoreoverinthecaseofmissingdatasamplesitusestheLeakyNoisy-orgatemodeltocalculatetheconditionalprobability．Thenitcanobtainamoreob-jectiveassessmentresult．ComparedwiththeassessmentofstaticBayesiannetworksdynamicBayesiannetworkmodelen-hancestheＲ＆Dprojectriskprobabilityassessmentaccuracywhichcanprovidesamorescientificbasisforriskcontrol．KeywordsprojectmanagementstaticBayesiannetworksdynamicBayesiannetworkＲ＆Dprojectriskprobability。StandishGroup28%1。BarryBoehm。ＲE=P×CPC2。13EM、4。。。。·131·。、、、、、、、、。LuigiPortinale5AnnMJulia6。、78。。LeakyNoisy-orgate。22.1X=X1X2…XnB＜BsBp＞BsXBpCPT。XPX1X2…Xn=∏ni=1PXi|PaXi1t10111。2。BoX11PXB→X1X2PXt+1Xtt。121213。1Bo2B→X0X1…XTPX1X2…XT=PB0X1∏Tt=1PB→Xt+1|Xt2PXtXt－1XititPaXitNNPXtXt－1=∏Ni=1PXitPaXit3PX1N1T=∏Ni=1PXi1|PaXi1×∏Tt=2∏Ni=1PB→Xit|PaXit42.2YHenrionLeakProbabilityLeakyNoisy－orgate14。LeakyNoisy－orgateYXiXallXallXiPiPallXlPY|Xi=Pi+Pall－PiPall5PY|Xi=Pall656PiPi=PY|Xi－PY|Xi1－PY|Xi7·231·4YP1…Pi…PnXLPLYPY=T=1－1－Pl∏iXi∈Xp1－Pi83。、、。。XX1、X2、X3。3。3。Genie1516。Genie。Bo4。4BoBoB→5。5B→state0=1state1=0。1007070/100=0.7。PX1=1=0.7PX2=1=0.6PX3=1=0.4。PX=1|X1=1=0.7PX=1|X2=1=0.8PX=1|X3=1=0.6PX=1|X1=0=0.5PX=1|X2=0=0.6PX=1|X3=0=0.4。8X1。1XXΦX1X2PX1=10.1000.4600.550PX1=00.9000.5400.450XX3X1X2X1X3PX1=10.4000.7300.640PX1=00.6000.2700.360XX2X3X1X2X3PX1=10.7000.820PX1=00.3000.180Bo·331·XPXT6。6PXTPX1=1=0.5PX2=1=0.4PX3=1=0.4。3PB→=0.22770.772347、89。7PX1T+18PX2T+19PX3T+1。B→10。10PXT+12。2State00.6070.4460.4280.356State10.3930.5540.5720.644411。113。·431·3State00.6070.5010.4760.372State10.3930.4990.5240.628。。。BS=1n∑ni=1Pi－Oj29PinOj10。0～101。。。state10.16960.2053。。5。。。LeakyNoisy－orgate。Gibbs。。1ProjectFailureandSuccessＲesearchＲeportＲ．StandishGroupInternationalChaos2005．2BoehmBW．SoftwareＲiskManagementPrinciplesandPracticesJ．SoftwareIEEE19918132－41．3．J．201046762－65．4．EM－GAJ．20102741360－1362．5LuigiPortinaleDanieleCodettaＲaiteriStefaniaMontani．Support-ingＲeliabilityEngineersinExploitingthePowerofDynamicBayes-ianNetworksJ．InternationalJournalofApproximateＲeasoning2010512179－195．6NicholsonAEFloresMJ．CombiningStateandTransitionModelswithDynamicBayesianNetworksJ．EcologicalModelling20112223555－566．7．J．20132712135－139．8．CVaＲJ．20132710111－115．9．J．201225526－30．10．M．2007.10．11．J．20093261－4．12DonatＲLerayPBouillautLetal．ADynamicBayesianNet-worktoＲepresentDiscreteDurationModelsJ．Neurocomputing2010734570－577．13．J．20101861856－1861．14Oni＇skoADruzdzelMJWasylukH．LearningBayesianNetworkParametersfromSmallDataSetsApplicationofNoisy-OＲGatesJ．InternationalJournalofApproximateＲeasoning2001272165－182．15．GeNIeJ．201124114－20．16．J．20098133－136．·531·