Evolutionary computation for automatic Web service composition: an indirect representation approach

EDA-based approach to comprehensive quality-aware automated semantic web service composition

“…Future work can investigate variable neighborhood with combinations of more than one local search operators in one evolutionary process, and investigate memetic EDA for handling multiobjective service composition problems. TABLE 2: Mean fitness values for our approach in comparison to NHM-EDA [12], PSO [10], MEGA [9] and GA. (Note: the higher the fitness the better) Dataset…”

“…We conduct experiments to evaluate the performances of our memetic EDA-based approaches, i.e., memetic EDA with constrained one-point swap (henceforth referred to as MEEDA-OP), memetic EDA with constrained two-point swap (henceforth referred to as MEEDA-TP), memetic EDA with constrained layer-based one-point swap (henceforth referred to as MEEDA-LOP) and memetic EDA with constrained one-block swap (henceforth referred to as MEEDA-OB). These memetic EDA-based approaches are compared to some state-of-the-art EC-based methods that were recently proposed to solve the same or similar problems: a PSO-based approach [10] (henceforth referred to as PSO), a GA-based approach (henceforth referred to as GA), a memetic GA-based approach [9] (henceforth referred to as MEGA) and an EDA-based approach [12] (henceforth referred to as NHM-EDA). Two benchmarks, WSC-08 [1] and WSC-09 [2] extended with QoS attributes , which generated from the QoS distribution from QWS [30] are created.…”

“…In the domain of service composition, to overcome the prematurity and proneness of GP, Tabu search is combined with GP to solve QoS-aware data-intensive web service composition [35]. [9] proposed an indirect memetic approach for QoSaware web service composition, where a domain-dependent crossover operator is proposed to produce candidate solutions. Besides that, an exhaustive local search is applied to composite solutions represented as permutations.…”

“…The one-point swap local search operator is inspired by [9], which swaps a pair of service indexes in a permutation. In [9], local search exclusively explores the neighborhood based on one selected index of the permutation, so the size of the neighborhood associated with the index is n − 1. However, it can be very computational expensive because the number of swaps becomes significant for large n. Besides that, it can be less flexible as the neighborhoods are just focusing on those neighborhoods in relation to one selected index.…”

“…EC techniques have been widely used to solve service composite problems that aim to optimize either one or both of QoSM and QoS, and are potentially more useful in practice as they can efficiently find "good enough" composite solutions. Important approaches [5], [6], [7], [8], [9], [10], [11], [12], [13] based on Genetic Algorithms (GA) [19], Genetic Programming (GP) [20], Particle Swarm Optimization (PSO) [21] and Estimation of Distribution Algorithm (EDA) [22], have been widely investigated in the literature.…”