Towards an efficient and a more accurate web service selection using MCDM methods

Nowadays, due to the fast growth of the number of web services with the same functionalities, the selection of the best services becomes a tricky task. Therefore, to provide a more powerful selection tool, web services are now additionally distinguished with respect to their non-functional aspects, especially to Quality of Service (QoS) requirements. Indeed, using QoS perspective in web service selection can help to promote the best candidate web services among those holding the same functional requirements. Hitherto, various QoS based web service selection approaches have been explored in the literature. However, two main challenging and conflicting issues are still demanding a specific attention, namely how accurate is the selection and what is the cost of such a process? In this paper, we target Multi Criteria Decision Making (MCDM) methods to select web services. Within this context, we propose a hybrid method to select the optimal web services based on MCDM. First of all, we use the Skyline method to reduce the research space by removing the dominated web services. Then, the BWM (Best Worst Method) is applied to normalize the weights associated with the QoS parameters according to the user request requirements. To rank the non-dominated web services with respect to the weights normalization, we explore the use of four different MCDM approaches that enjoy an affordable complexity, which are: VIKOR, SAW, TOPSIS and COPRAS. For the sake of a comparison, each method is first tested separately in order to assess its ability to select the optimal elements. Then, in order to provide the best compromise between the different ranking methods, we apply the Borda voting method to determine the compromise solution. Hence, we are able to measure the accuracy of each technique relatively to the Borda solution by applying a similarity function. The implementation and the simulation of the different methods have permitted to assess their performances in terms of time execution and accuracy. The results show clearly that MCDM ranking methods are more likely to diverge in selecting the best elements when the number of web services is increased. This finding argues for the pertinence of computing the Borda solution in the context of MCDM based web service selection.