Evolving support vector machine parameters

The kernel type, kernel parameters and upper bound C control the generalization of support vector machines. The best choice of kernel or C depends on each other and the art of researchers. This paper presents a general optimization problem of support vector machine parameters including a mixed kernel and different upper bounds for unbalanced data. The objectives are ξa-estimators of the error rate, recall and precision. Evolutionary algorithms are used to solve the problem. The performance of this method is illustrated with a standard data set of intrusion detection application.