Combining Von Neumann Neighborhood Topology with Approximate-Mapping Local Search for ABC-Based Service Composition

Abstract

Service composition with end-to-end QoS constraints have been proven to be an NP-hard problem and various evolutionary algorithms such as Artificial Bee Colony (ABC) are widely adopted to look for an approximately-optimal solution in the restricted time. The advantage of ABC algorithm is its simplicity (i.e., only three control parameters, and simple heuristic rules for exploiting the solution space), and our previous work has verified its effectiveness in solving the service composition problem. This paper focuses on the enhancement of traditional ABC neighborhood strategy for local search, with the objective of better optimality and faster convergence rate. The work is shown in two perspectives. Firstly, an approximate-mapping based local search strategy is proposed, where the discrete solution space of service composition problems are approximately transformed into a continuous space in which a locally optimal neighboring solution is precisely found, in this way, the superiority of traditional ABC could still hold in service composition problem. Secondly, we adopt the Von Neumann neighborhood topology, which has been proven to have better performance than other topologies, to further improve the quality of local search. Experiment results show that our Approximate-Mapping Von Neumann algorithm (AMV) is more effective than other service composition algorithms such as genetic algorithm and Threshold-Based Algorithm (TBA).