Research On Interval Adaptive Particle Swarm Optimization And Its Application

Particle Swarm Optimization (PSO) is a novel swarm intelligence technique, which was inspired by swarm intelligence of bird flocking and fish schooling and developed by Eberhart and Kennedy in 1995. The characteristics of PSO are simple and only a few parameters needed to be adjusted, easy implementation and fast convergence. Because of these advantages, PSO is more and more popular among the researchers. At present, a lot of people are doing research on the Particle Swarm Optimization algorithm, and apply it on the function optimization, neural network training, fuzzy system control, pattern recognition and other fields. It has a good prospect of application.The basic principle, parameters, several kinds of improved PSO and the application in the Intrusion Detection System are discussed in this paper; the main details are as follows:Detailed analyze the influence of inertia weight factor, shrinking factors, population size and the topological structure on PSO algorithm, and experiments of inertia weight factor show that, the PSO algorithm with nonlinear inertia weight factor has a better effect and strong adaptability. In addition, the theories of PSO algorithm are discussed, including the space form, the convergence of mathematical analysis and convergence, etc, and the simulation results reveal that, in a certain extent, the single particle trajectory reflected the particle swarm trend.Basic particle swarm algorithm cannot solve the problem with discrete variables, and it will have the problem of early maturing because of the easily convergence. In view of this situation, this article focuses on several improvement methods of the particle swarm optimization algorithm:discrete PSO, chaotic PSO, simulated annealing PSO and immune PSO. However, in the aspect of handling large amounts of data, the speed and accuracy that PSO optimized are slow. So the paper puts forward a new kind of interval adaptive particle swarm algorithm, in which the inertia weight factor changes as the situation varies. After using the interval adaptive particle swarm algorithm optimizes the parameters of the SVMs, we use the SVM to classify the intrusion data of the Intrusion Detection System. The experiment results showed that this method improves the classification accuracy and shortens the response time. That proves this method is workable.