Reliable Service Composition via Automatic QoS Prediction

Service composition has received considerable attention nowadays as a key technology to deliver desired business logics by directly aggregating existing Web services. Considering the dynamic and autonomous nature of Web services, building high-quality software systems by composing third-party services faces novel challenges. As a solution, new techniques have been recently developed to automatically predict the QoS of services in a future time and the prediction result will facilitate in selecting individual services. Nonetheless, limited effort has been devoted to QoS prediction for service composition. To fill out this technical gap, we propose a novel model in this paper that integrates QoS prediction with service composition. The integrated model will lead to a composition result that is not only able to fulfill user requirement during the composition time but also expected to maintain the desired QoS in future. As user requirement is expected to be satisfied by the composition result for a long period of time, significant effort can be reduced for re-composing newly selected services, which usually incurs high cost. We conduct experiments on both real and synthetic QoS datasets to demonstrate the effectiveness of the proposed approach.