The Research of Support Vector Machine with Optimized Parameters Based on Particle Swarm Optimization

SVM (Support Vector Machine), which is based on structural risk minimum principle, overcomes the shortness of the traditional machine learning method, especially fit for the small sample problem, it is the focus of the failure diagnose field. But at present there is not a definite theory to guide the choice of SVM parameters. It has great influence on the classification performance and operating speed to choose the proper parameters of SVM. In this paper, Particle Swarm Optimization algorithm is utilized to optimize the parameters of SVM and its kernel function to improve the performance by properly choosing the fitness function. Simulation result demonstrates that this method can greatly improve the over-all properties of the failure diagnose system. The optimizaton result is better than Genetic algorithm(GA).