基于LMD改进GPR优化的网络流量预测

Network traffic prediction based on LMD improved GPR optimization

针对蚁群算法容易陷入局部最优、网络流量预测准确性不高的问题,提出一种基于局部均值分解(LMD)改进蚁群优化高斯过程回归(GPR)的预测算法。考虑到网络流量的复杂性,使用LMD将网络流量分解成多个相关的子序列;通过GPR对网络流量子序列进行建模分析;用蚁群算法优化超参数,引入视线角度参数控制蚂蚁搜索时的视线范围,提高蚂蚁的局部搜索能力;通过莱维飞行更新蚂蚁搜索的步长,提高蚁群算法搜索的全局性。实验表明,改进后的蚁群算法搜索到了更优的值,与原有GPR算法相比,LMD分解后改进蚁群优化GPR的算法来预测网络流量,更好地拟合了网络流量的走向,提高了预测的效果,对维护网络安全具有一定的作用。

It is easy for ant colony algorithm to fall into local optimum and poor accuracy in network traffic prediction.An improved ant colony optimization GPR prediction algorithm based on LMD is proposesd.Firstly,according to the complexity of network traffic,local mean decomposition(LMD)is used to decompose the network traffic into multiple related subsequences.Secondly,Gaussian process regression(GPR)is applied to model and analyze the network traffic subsequences.Ant colony algorithm is used to optimize the super parameters.Thirdly,by introducing the line of sight angle parameter to control the line of sight of ants in search,the local search ability of ants is improved.Fourthly,the step size of ant colony algorithm is updated by Levy flight to improve the global search.Experimental results show that the improved ant colony algorithm finds a better value.Compared with the original GPR algorithm,the improved ant colony optimization GPR algorithm after LMD decomposition can predict network traffic,better serve the trend of network traffic and improve the prediction effect,in a maintaining network security role.