Skip to Main Content
Support vector machine for pattern classification is motivated by linear machines, but rely on preprocessing the data to represent in a high dimension with an appropriate nonlinear mapping, data from two categories can by separated by a hyperplane. To make certain the hyperplane, the key problem is selecting appropriate criterion and algorithm. To find out the appropriate solution vector in solution spaces, fixed increment, variable increment, relaxation, and stochastic approximation etc. may be selected, this article provide a novel method-modified general particle swarm optimization for finding the solution vector. The proposed method enhances performance and avoid over fitness effectively.