基于自适应模糊神经USP的面向对象的软件工作量估计(IJISA-V4-N6-2)

I.J.IntelligentSystemsandApplications,2012,6,14-24PublishedOnlineJune2012inMECS()DOI:10.5815/ijisa.2012.06.02Copyright©2012MECSI.J.IntelligentSystemsandApplications,2012,6,14-24ObjectOrientedSoftwareEffortEstimatewithAdaptiveNeuroFuzzyuseCaseSizePoint(ANFUSP)MohammadSaberIrajiDepartmentofComputerscience,YoungResearchersClubsariBranch,IslamicAzadUniversity,sari,Iran.E-mail:[e-mail]mentofComputerengineering,IslamicAzadUniversity,SariBranch,Sari,IranE-mail:[e-mail]—Usecasesizepoint(USP)methodhasbeenproposedtoestimateobjectorientedsoftwaredevelopmenteffortinearlyphaseofsoftwareprojectandusedinalotofsoftwareorganizations.Intuitively,USPismeasuredbycountingthenumberofactors,preconditions,postconditions,scenariosincludedinusecasemodels.InthispaperhavepresentedaAdaptivefuzzyNeuralNetworkmodeltoestimatetheeffortofobjectorientedsoftwareusingUseCasesizePointapproach.Inourproposedsystemadaptiveneuralnetworkfuzzyusecasesizepointhaslesserrorandsystemworkedmoreaccurateandappropriativethanpriormethods.IndexTerms—Usecasesizepoint,effort,adaptiveneurofuzzy,objectorientedeffortI.IntroductionTodaythesoftwareindustryexperienceandtrainingdatainordertoestimatetheeffortofsoftwareisimportant.Estimationofobjectorientedsoftwarecostandeffortisanimportantandhardmanagementactivity.Thisisduetothelackofinformationtomakingdecisionsintheearlyphasesofthedevelopment,frequentlycharacterizedbyuncertainty.Tohelpthemanagersinthistask,thereareintheliteraturemanyestimationmodelsthatusuallyincludetwomainmetrics:LinesofCode(LOC)andFunctionPoints(FP)[2],bothofthemwithskillsandlimitations.LOCisdependentontheprogramminglanguageandtheFPAnalysis(FPA)issubjectiveandbasedonhumandecisions[3].Asthemostpopulartechniqueforobject-orientedsoftwarecostestimation,UseCasePoints(UCP)method,however,hastwomajordrawbacks:theuncertaintyofthecostfactorsandtheabruptclassification[4].Softwarecostestimationisvitalforprojectbidding,budgeting,controllingandplanning.AlthoughintheliteraturemanyestimationmodelslikeConstructiveCostModel(COCOMO),FunctionPoints(FP)havebeenproposedtohelpmanagerinestimationtask,thereisnoobviousevidenceshowsthattheaccuracyisimprovedinlastdecades[5].Achievingahighlyaccurateestimationisstillachallengingissueinsoftwareengineering[4].Thesubjectsofestimationintheareaofsoftwaredevelopmentaresize,effortinvested,developmenttime,technologyusedandquality.Particularly,developmenteffortisthemostimportantissue.Sofar,severaleffortmodels[6][7][8]havebeenproposedandmostofthemincludesoftwaresizeasanimportantparameter.Functionpointisameasureofsoftwaresizethatuseslogicalfunctionaltermsbusinessownersandusersmorereadilyunderstand[9].Sinceitmeasuresthefunctionalrequirements,themeasuredsizestaysconstantdespitetheprogramminglanguage,designtechnology,ordevelopmentskillsinvolved.Toestimatetheeffortinearlierphase,usecasepointmethodhasbeenproposed[10].Usecasepoint(UCP)ismeasuredfromausecasemodelthatdevelopsthefunctionalscopeofthesoftwaresystemtobedeveloped.Itisintendedbythefunctionpointmethodsandisbasedonanalogoususecasepoint.Thereareseveralexperiencereportsthatshowtheusefulnessofusecasepointforearlyestimationofsoftwaresize.UnifiedModelingLanguage(UML)isagraphicalmodelinglanguagethatisusedforvisualizing,specifying,constructinganddocumentingsoftwaresystems.Tocapturethefunctionalrequirementsofasoftwareprojectusecasemodelsareoftenemployed.Usecasemodelingisatechniquethathasbeenwidelyusedthroughouttheindustry/researchtodescribeandcapturethefunctionalrequirementsofasoftwaresystem[11].Sinceusecasesandscenariosaredevelopedasanormalpartofrequirementsgatheringandanalysistheycaptureanaccuraterepresentationoftheuser‘srequirement.Theusecasepoint‘smethodadjuststhesizeofthefunctionalityofthesystembasedonanumberoftechnicalandenvironmentalfactors.Thetechnicalfactorsarerelatedtonon-functionalrequirementsonthesystem,whiletheenvironmentalfactorscharacterizetheObjectOrientedSoftwareEffortEstimatewithAdaptiveNeuroFuzzyuseCaseSizePoint(ANFUSP)15Copyright©2012MECSI.J.IntelligentSystemsandApplications,2012,6,14-24developmentteamanditsenvironment.Theinfluenceofthesefactorsontheestimatewasinthiscasemuchsmaller(a16%increaseintheestimate)thantheincreaseinactualeffortspentbythecompaniesthatemphasizedthedevelopmentprocessandthequalityofthecode(anincreaseinactualeffortofmorethan100%)[1].Softwareestimationmodelscombiningalgorithmicmodelswithmachinelearningapproaches,suchasneuralnetworksandfuzzylogic,havebeenviewedwithskepticismbythemajorityofsoftwaremanagers[12].Briefly,neuralnetworktechniquesarebasedontheprincipleoflearningfromhistoricaldata,whereasfuzzylogicisamethodusedtomakerationaldecisionsinanenvironmentofuncertaintyandvagueness.However,fuzzylogicalonedoesnotenablelearningfromthehistoricaldatabaseofsoftwareprojects.Oncetheconceptoffuzzylogicisincorporatedintotheneuralnetwork,theresultisaneuro-fuzzysystemthatcombinestheadvantagesofbothtechniques[13].Manymetricsexistbasedonusecasemodelinobjectorientedsoftware.weConsiderusecasesize