摘要
为了提高网络流量预测的精度,针对网络流量数据具有非线性、非平稳的特点,提出一种基于经验模态分解(EMD)和混沌粒子群算法优化组合核极限学习机的网络流量预测模型。首先将网络流量时间序列进行EMD分解,提取网络流量数据的各个分量,然后分别对各个分量采用核极限学习机进行预测,最后重构出预测结果。针对传统核极限学习机拟合能力的不足,提出一种基于高斯核和多项式核组合的组合核极限学习机,并且采用改进的混沌粒子群算法优化组合核的核参数组合权值以及惩罚因子,并将其应用到网络流量预测中。实验结果表明,该方法可以有效提高网络流量预测的精度,有助于指导网络资源的合理分配和规划。
In order to improve precision of network flow prediction, a prediction model is proposed in this paper based on Empirical Mode Decomposition （EMD） and chaos particle swarm optimization combined kernel extreme learning machine aiming at the features of nonlinear and non-stationary for network flow data. Unit flow is obtained through EMD on the network flow in time sequence, then each unit data is predicted with kernel extreme learning machine. Finally ,the prediction result is reconstructed. In view of the inadequate fitting capacity of traditional kernel extreme learning machine, a machine combining Gaussian kernel and multinomial kernel is proposed and the improved kernel parameter combination and penalty factor of chaos particle swarm optimization with combined kernel are applied in the prediction of network flow. The experiment shows that this method can improve the accuracy of network prediction effectively, and help guide the rational allocation and planning of network resources.
出处
《计算机技术与发展》
2016年第6期73-77,共5页
Computer Technology and Development
基金
河南省教育科学技术研究重点项目(15C520016)
开封市科技攻关计划项目(130145)
关键词
网络流量预测
核极限学习机
组合核函数
混沌粒子群
经验模态分解
network flow prediction
kernel extreme learning machine
combined kernel function
chaos particle swarm
empirical mode decomposition
