Seite - 362 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

to theWeighting Agentasanasynchronousmessage.ThisAgentweights theservices andsendsanasynchronousmessagetotheRanking Agent toclassifythembasedonthe calculatedscores.ThebestserviceissentasanasynchronousmessagetotheRequester Agent. 4. WeightingofQoSCriteriausingAHP Despite, thereareanumberofmethodsforassigningweights to thecriteria,but theAHP is themost used[22]. In this section,wepresent howwehave applied theAHPmethod to designateweights related to quality criteria. First, the quality of service is defined in ourworkasacombinationofseveralcriteriawhichmaybequantitativesuchasresponse time, availability, reliability, etc. orqualitative suchasusability, accessibility, etc.These parameters can thenbeconsideredas criteria of choicebetween several discoveredweb services that respect theclient requirements. In thiswork,wehaveselectedfourcriteria: • Reliability: is the ability of a service to perform the required functions under specifiedconditions for agivenperiodof time[23]. • Throughput: is thenumberofcompletedservicerequestsoveratimeperiod[24]. • Responsetime: is the timedurationbetweenaserviceusersendingarequestand receivinga response[25]. • Cost: is theprice tobe spent toexecuteawebservice, this costmaybeprovided by the serviceprovider[5]. TheAHPuses a set of numerical rates,where each rate is related to agivenprefer- ence (seeTable1). Table1. PreferenceScale[19] VerbalJudgmentofPreference NumericalRating Extremelypreferred 9 Verystrongly toextremelypreferred 8 Verystronglypreferred 7 Strongly toverystronglypreferred 6 Stronglypreferred 5 Moderately to stronglypreferred 4 Moderatelypreferred 3 Equally tomoderatelypreferred 2 Equallypreferred 1 The following three-step procedure provides a good approximation of the synthe- sizedpriorities[19]: • Step1:Sumthevalues ineachcolumnof thepairwisecomparisonmatrix. • Step2:Divideeachelement in thepairwisematrixby its column total. • Step3:Compute the averageof the elements in each rowof thenormalizedma- trix. H.Belouaaretal. /AHPandTOPSISMethods forWebServiceSelection362