Parameter selection for support vector machines based on hybrid genetic algorithms

The general parameter selection method for support vector machines based on the steepest descent algorithm tends to trap into the local minimum point. To overcome this shortcoming, a parameter selection method based on hybrid genetic algorithms is proposed. The proposed method employs the global and local optimization capability brought by genetic algorithms and the steepest descent algorithm, thus better parameters for support vector machines can be determined. Experimental results demonstrate an improvement of the generalization performance for support vector machines.