An Empirical Study on Meta-Heuristic Search-Based Web Service Composition

In recent years, quite a few search algorithms have been used to solve Web service composition problem. However, it is still lack of systematic analysis about these methods. In the paper, we attempt to analyze the effect of three typical meta-heuristic search algorithms. The experimental analysis is performed according to the search-based Web service composition framework. In the experiments, we mainly carry out the analysis as follows: different abstract service number, different candidate service number and different QoS constraint strength. Based on the above analysis, some guidelines of realizing Web service composition via search algorithms are yielded. For example, EDA is suitable for the composition problem with large abstract service number. PSO is more effective for the case of large candidate service number. GA is a good choose for the little-scale Web service composition problem. In addition, EDA can effectively settle the composition problem with strong QoS constraint. GA is suitable for the problem with medium constraint strength, and PSO has the poor ability to tackle the QoS constraint with high strength.