How to deal with QoS value constraints in MCDM based Web service selection

Abstract

SummaryThe aim of the current research is to develop a more accurate Web service selection approach that can deal with value constraints on QoS criteria in the user request. The purpose is to be able to promote services even when all the value constraints are not satisfied and hence rank them according to their closeness to meet the latter. For this effect, we use MCDM (Multi Criteria Decision making) methods together with suitable normalization techniques at different stages of the process. First of all, we consider an extension of the AHP method to compute and normalize the weights associated with the QoS criteria considered in the Web service selection. Second, we introduce a more consistent normalization technique, called OMRI, to normalize data according to value constraints. Third, we propose to extend different ranking MCDM methods to the OMRI normalization so that it becomes possible for them to cope with value constraints. The considered methods are SAW, TOPSIS, VIKOR, and WPM. To compare the accuracy of the extended ranking methods, we use the Borda compromise solution together with different similarity ratios. To validate the solution, several experiments have been conducted on a real dataset as well as on an artificial one.