An Empirical Assessment on Program Module-Order Models

摘要

Softwarereliabilitycanbeimprovedbydetectinglatentsoftwaredefectspriortodeployment.Apracticalapproachistotargetavailableresourcestowardprogrammoduleslikelytohavefaults,i.e.,low-reliabilitymodules.Anempiricalinvestigationonprogrammodule-ordermodels(MOM)ispresented.Amodule-ordermodelassistsmanagementinprioritizingverificationandvalidationtaskstowarddefectivemodules,providingagoal-orientedsoftwarereliabilityenhancement.TheoutputofaMOMisaquality-basedrankingofprogrammoduleswherequalityisestimatedbyanunderlyingsoftwaredefectpredictionmodel.Wepresentanempiricalinvestigationonmodule-ordermodelsdesignedusingdifferentunderlyingdefectpredictionmodels.Basedoncasestudiesofvariousindustrialsystems,itisobservedthatthereisnosinglefaultpredictionmodelthatalwaysresultsinthepreferredMOM.Infact,performanceofaMOMwasdependentonthecutoffrangeofinterest,i.e.,howmanymodulescanbeinspectedwiththegivenresources.Consequently,apractitionercanselectaMOMbasedonagivencutoffrangeofinterest.