Research On Improved Strategy Of Multi-objective Particle Swarm Optimization

Multi-objective optimization plays an important role in solving problems with multiple conflicting goals.It has a wide range of applications in real life.Particle swarm optimization is a swarm intelligence algorithm developed by simulating bird foraging.Compared with other swarm intelligence algorithms,particle swarm optimization has advantages like few parameters and fast convergence speed in solving multi-objective optimization problems,but it also has disadvantages like falling into local best easily.Based on consulting a large amount of literature and the systematic research on multi-objective particle swarm optimization,the multi-objective particle swarm optimization's distributed performance and convergence performance are improved,and the improved algorithm is applied to optimization problems in Web service composition.The main work of this paper is focused on the following aspects:1.A coarse-grained model using hierarchical subgroups to improve the performance of particle swarm optimization was proposed.The particle swarm was divided into several sub groups in the evolutionary process,different subgroups played different the role in the evolutionary process,every subgroups was evolved indenpendently,the information was exchanged between subgroups after certain iteration times to improve the search breadth of particles.2.A dynamic adjustment strategy was proposed to improve the distributed performance and convergence performance of particle swarm optimization in solving multi-objective optimization problems.the external Pareto solutions set was detected in the process of evolution to assess the evolutionary state,the suitable evolution strategy was adjusted according to different evolutionary state to ensure that the algorithm could consider distribution performance and convergence performance in the process of evolution.3.For the Web service combination optimization problem,the multi-objective optimization problem model that execution time and service cost played as the optimization objective and reputation rating and reliability played as the constraint conditions was created,and this multi-objective optimization problem model was solved by using two kinds of optimization strategies.through comparative analysis,the two strategies' feasibility and effectiveness were verified.