Aiming at the support vector machine(SVM) kernel parameters selection problem, an optimal selection method is proposed based on Fisher criterion and maximum entropy principle. A criterion for measuring the performance of the SVM kernel parameters is proposed by analyzing the classification principle of SVM. According to this criterion, the Fisher criterion is used to optimizing the SVM kernel parameters. Then, the maximum entropy principle is introduced to adjust the optimization performance of the algorithm. Finally, the whole model uses the particle swarm optimation (PSO) algorithm for parameters optimization. The experiment results on datasets from UCI show that the proposed method is more apt on parameter selection and can improve the generalization performance of SVM.