Optimising the semantic web service composition process using bio-inspired methods

Abstract

This paper presents two bio-inspired methods (one inspired by the cuckoo’s breeding behaviour, and another one inspired by natural evolution and genetics) for selecting the optimal or near-optimal solution in web service composition. The proposed methods are applied on an enhanced planning graph structure which models the composition search space for a given user request. The cuckoo-inspired selection method applies a 1-OPT heuristic to expand the search space in a controlled way such that the stagnation in a local optimum solution is avoided. The genetic-based selection method uses two memory structures to avoid the stagnation in a local optimum solution on one hand, and to ensure that exploitation and exploration are properly performed. The quality of a composition solution is evaluated in terms of QoS attributes and semantic quality. To validate the proposed methods we have implemented an experimental prototype and carried out experiments on a set of scenarios with different complexities. Finally, we comparatively analyse the experimental results obtained by applying the two selection methods.