简介：Forworkflow-basedservicecompositionapproach,therelationsbetweentheWebserviceQoSandenvironmentsareusuallynotconsidered,sothattheinformationaboutQoSforcompositeserviceselectionisinaccurate.Itmakestheselectedcompositeserviceinefficient,orevenunexecutable.Toaddressthisproblem,anovelservicecompositionapproachbasedonproductionQoSrulesisproposedinthispaper.Generally,itisverydifficulttodirectlyanalyzehowdifferentkindsofenvironmentfactorsinfluencetheWebserviceQoS.Weadopt'black-box'analysismethodofoptimizingcompositeservices,discoveringtheknowledgesuchas'theQoSofoneWebservicewillbehigherinspecificenvironments'.Inourapproach,theexecutioninformationofthecompositeserviceisrecordedintoalogfirst,whichwillbetakenasthebasisofthesubsequentstatisticalanalysisanddatamining.Then,thetimelyQoSvaluesoftheWebservicesareestimatedandtheproductionQoSrulesbeingusedtoqualitativelyexpressthedifferentperformancesoftheWebserviceQoSindifferentenvironmentsaremined.Atlast,weemploytheminedQoSknowledgeoftheWebservicestooptimizethecompositeserviceselection.Extensiveexperimentalresultsshowthatourapproachcanimprovetheperformanceofselectedcompositeservicesonthepremiseofassuringtheselectingcomputationcost.