Study On Optimizing Service Composition Based On Improved Imperialist Competitive Algorithm

With the rapid development of computer technology, Web, Grid, Cloud and other new technologies quickly mature and gradually integrate into people's lives, and Service is the core of them. A single service can only provide limited function, when the user's specific needs can not be meet by a single service provided by the network, we need to use the Service Composition technology to form a service by some means based on existing services to meet the user's needs.With the emergence of a large number of Web Services with the same functional attributes and different non-functional attributes,how suitable service instances can be dynamically selected from a large number of candidate services, thus forming an executable composite service with global optimal quality of service(QoS) that can meet user needs becomes a key problem in Service Composition,we called it Service Composition Optimization Problem in this paper. Traditional Web Service Selection is generally more concerned with the functional requirements of service, while ignoring the non-functional indicators of Composed Service.At present, swarm intelligence optimization algorithm, such as Particle Swarm Optimization, Ant Colony Optimization and so on are widely studied and used on Service Composition Optimization Problem based on QoS, compared to the Particle Swarm Optimization, Ant Colony Optimization and other swarm intelligence optimization algorithms, Imperialist Competitive Algorithm(ICA) is more novel and universal, and it has excellent performance in the optimization problem, which is more suitable for solving the Service Composition Optimization Problem.In this paper, we first establish the QoS model of the service, and give a method to aggregate the QoS attribute value of the four basic combination models, and further through the iterative method to obtain the Qo S attribute value of complex service composition model. On this basis, this paper constructs a representative abstract service composition model, and uses the Imperialist Competitive Algorithm to solve the Service Composition Optimization Problem on this model. In order to verify the effect of this method, we use the Particle Swarm Optimization which is most often used currently in Service Combination Optimization Problems as the control algorithm, through the contrast experiment, the feasibility and effectiveness of Imperialist Competitive Algorithm in solving the problem of Service Composition is verified.Finally, we improved the original Imperialist Competitive Algorithm, and proposed a new Improved Imperialist Competitive Algorithm. This algorithm improves the performance of the ICA algorithm by improving the initial state generation strategy, the introduction of the splitting operator, and the fusion of Gravitational Search Algorithm etc., and experimental results demonstrate that the Improved Imperialist Competitive Algorithm outperforms the original Imperialist Competitive Algorithm in three aspects, the fitness, the optimization efficiency and the stability.