Eliminating the correlations of quality of service criteria in web service selection

QoS (Quality of Service) is an important aspect to measure the running performance of a web service. Through integrating the values of various QoS criteria, we can evaluate the overall quality of a web service. However, the traditional QoS-aware service selection methods, e.g., SAW (Simple Additive Weighting) often assume that all the QoS criteria are independent of each other. While according to our observation on WSDREAM (a real-world web service QoS Dataset), the published QoS criteria are highly correlative, not independent. In this situation, it's a great challenge for traditional service selection methods to deal with the correlated QoS criteria. In view of this challenge, in this paper, Mahalanobis distance is recruited to eliminate the correlations among various QoS criteria, based on which we put forward a novel Mahalanobis distance-based service selection method, i.e., MD_SSM (Mahalanobis Distance-based Service Selection Method, MD_SSM). Finally, a set of experiments are deployed to validate the feasibility of our proposal, and the experiment results show that MD_SSM can work in most service evaluation situations where QoS criteria are correlative.