Dealing with user constraints in MCDM based web service selection

Abstract

Nowadays, web service selection has become a very challenging issue. As the number of web services is increasing drastically, different QoS parameters are added in their description in addition to their functional parameters (Inputs and outputs) to distinguish them and ease their selection. Hence, web service selection consists to find out among services having the same functionalities the optimal elements that satisfy user QoS requirements. When the user request deals with value constrains on the QoS parameters, the mainstream approach operates by removing from the search space web services that do not satisfy the user constraints. However, such an approach is not always appropriate especially when the resulting search space is empty or very small. In this case, the discarded services may be of interest in the selection process. To tackle this issue, we propose to use a combination of two MCDM (Multi Criteria Decision Making) methods, that are AHP (Analytical Hierarchy Process) and RIM (Reference Ideal Method). AHP is used to encode an normalize weights constraints assigned to QoS criteria at the first stage. Then to deal with user value constraints, we consider RIM to rank discarded web services relatively to their proximity to the required value constraints. We compare our method with existing approaches by performing tests on a real dataset.