SVM Algorithm Based on Interval Adaptive PSO and Its Application

The parameters of the kernel functions are very improtant for the SVMs(Support Vector Machines) in the generalization ability,especially when we use the SVM for data classification with a great deal of data,it will need so much computer memory that the speed of parameter optimization will be decreased.For this problem,this paper presents a method that uses an interval adaptive particle swarm optimization to optimize the parameters of the SVMs.Then we apply this method to the intrusion detection systems,and compare it with the Ant Colony Algorithm and the Genetic Algorithm.The experimental results show that this method improves the classification accuracy by 9.7%,and the response time is shortened by 40.6%～56.5%.That proves this method is workable.