A method to choose kernel function and its parameters for support vector machines

The support vector machines are the new statistical learning algorithm which is developed in recent years. They have some advantages in many regions like pattern recognition. The kernel function is important to its classification ability. This paper presents a crossbreed genetic algorithm based method to choose the kernel function and its parameters. The crossbreed genetic algorithm uses two fitness functions which are produced according to the two criterion of SVM´s performance. The experiments proved that this algorithm can find effectively the optimal kernel function and its parameters, and it is helpful to increase the support vector machines´ performance in fact.