A Solution Space Reduction Approach based on Neural Network and Clustering for Large-scale Service Composition

Abstract

Service composition is an important way to generate value-added services in cloud computing. The selection of optimal composition scheme for QoS-aware service composition has become a research hot topic. In a dynamic cloud service environment, the complex service composition process structure and numerous candidate services generate a huge number of service composition results. It is challenging to select best or near best service composition result that meets the user's preference in a large number of solution spaces. We propose a service composition method based on neural network model prediction and clustering to ensure the optimality of service combination results based on user preferences while further reducing the service composition algorithm search time. We carry out comparative experiments of different approaches to validate the superiority of our approach in the background of large-scale service composition.