Knowledge-Driven Automated Web Service Composition - An EDA-Based Approach

Service Oriented Architecture starts with the concept of web services, which give birth to an application of web service composition that selects and combines web services to accommodate users’ complex requirements. These requirements often cover functional parts (i.e., semantic matchmaking of services’ inputs and outputs) and non-functional parts (i.e., Quality of Service). Service composition is an NP-hard problem. Evolutionary Computation (EC) techniques have been successfully proposed for finding solutions with near-optimal Quality of Semantic Matchmaking (QoSM) and/or Quality of Service (QoS) using knowledge of promising solutions. Estimation of Distribution Algorithm (EDA) has been applied to semi-automated QoS-aware service composition, since it is capable of extracting knowledge of good solutions into a explicit probabilistic model. However, existing works do not support extracting knowledge for fully automated service composition that does not obeying a given workflow. In this paper, we proposed an EDA-based fully automated service composition approach to jointly optimize Quality of Semantic Matchmaking and Quality of Services. This approach is compared with a PSO-based approach that was recently proposed to solve the same problem.