【摘要】 为更加精确地对服务器性能进行评估与预测,提出一种基于差分进化(DE)与灰狼寻优(GWO)相结合的SVM模型(DE-GWO-SVM)。利用灰狼寻优算法(GWO)寻求SVM的最优参数组合惩罚因子C和核函数参数γ,提升SVM算法的预测性能,将DE算法用于生成灰狼寻优算法初始种群的最优值,克服GWO的初始种群随机生成的局限性,使GWO具有更加良好的寻优能力,获取SVM算法的参数组合C和γ的最优解。实验结果表明,相比于传统的SVM、ABC_SVM、GWO_SVM模型,DE_GWO_SVM预测模型具有较高的预测精度、良好的稳定性和泛化能力。更多还原

【Abstract】 To evaluate and predict server performance more accurately,a SVM algorithm(DE-GWO-SVM)based on differential evolution(DE)and wolf hunting(GWO)was proposed.Gray wolf algorithm(GWO)was used to seek the optimal parameters of combined penalty factor C and the kernel function parameterγof SVM,which improved the prediction performance of the SVM algorithm,the DE algorithm was used to generate the optimal value of initial population of the gray wolf algorithm,which overcame the limitations of the initial population of GWO that it was randomly generated,which made GWO have greater optimization ability,and obtain the optimal solution of combined parameters C andγof the SVM algorithm.Experimental results show that compared with the traditional model of SVM,ABC_SVM and GWO_SVM,the model of DE_GWO_SVM has higher prediction accuracy,better stability and generalization ability.更多还原