A predictive intelligent method for optimization of web service composition

In the current high-speed computing environment, the number of web services is continuously increasing and massive service data is generated all the time. Web service composition can bring new challenges that cannot be accomplished by human work. The model of web service composition consists of multiple service nodes (a collection of web atomic services), and each service node contains multiple functionally similar web concrete services, but with distinctive quality of service (QoS) attribute values. The selection of specific services that satisfy concrete QoS constraints from each service node and formation an executable service combination to meet the user's service request is a NP-hard problem. A new difficulty in this field involves predicting the result of the composition and then achieving a global optimization. In this pursuit, the present study examined a new measure of service composition prediction by devising a new high-performance algorithm named PSO-GA (The hybrid optimization algorithm of Particle swarm optimization and Genetic algorithm), based on the advantages of particle swarm optimization algorithm and genetic algorithm. We developed the preservation of the local and global optima and the evolution of the two optimal directions within the particle swarm algorithm, and we added crossover and mutation operations to the PSO algorithm. An elite retention strategy was utilized to improve the convergence productivity of the calculation. It compensated for the standard particle swarm algorithm that was prone to the local convergence, and optimized the complex calculations as well as the slow evolution speed of the genetic algorithm. Experimental results showed that, compared with the GA and PSO, the algorithm proposed in this paper solved the optimization problem of service combination model with 17.6% and 16% better optimization accuracy, respectively, while the optimization time efficiency was improved by 32.3% and 13.5%, respectively. Our results indicated that the proposed intelligent hybrid algorithm performed well in the web service combination optimization model, and can demonstrate an increased service prediction potential in meeting the challenges of the future data explosion.