Network burst traffic changes have gradually substituted the smooth voice service,and the network needs to be able to dynamically predict and adjust the flow distribution.AR(n) model,a network traffic prediction algorithm,has several shortages compared with self-similar model on prediction accuracy,but holds well computing performance.Particle swarm optimization algorithm is adopted to optimize the AR(n) model with the AIC criterion.The experimental results on an open dataset show that the improved algorithm acquires well prediction accuracy compared with least-square,grey theory and self-similar models.%网络突发式的流量变化已经逐步取代平稳的语音服务,需要网络能够动态预测和调整流量分配.AR(n)模型作为网络流量预测中常用算法,相比于自相似模型其预测的精度稍有不足,但在计算性能上表现较好,以粒子群算法对AR(n)模型进行预测精度调优,采用AIC准则判断模型的最佳阶数.通过实验比较研究了最小二乘、灰色理论估计和自相似模型在预测精度和计算性能上差异性,实验结果表明,基于粒子群的AR(n)模型具有较好的预测精度.
展开▼