改进极限学习机的网络流量混沌预测  

Chaotic prediction of network traffic based on improved extreme learning machine

作　　者：刘蕴[1] 焦妍 王华东[3] 

机构地区：[1]周口职业技术学院信息工程学院,河南周口466000 [2]河南应用技术职业学院信息工程学院,河南开封475000 [3]周口师范学院计算机科学与技术学院,河南周口466001

出　　处：《南京理工大学学报》2017年第4期454-459,共6页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(U1504613);河南省高校科技创新团队计划(17IRTSTHN009)

摘　　要：为了获得更加精确的网络流量预测,降低网络拥塞的频率,提出了改进极限学习机的网络流量预测模型。针对网络流量混沌性分别确定原始网络流量的延迟时间和嵌入维数,采用极限学习机对网络流量的变化特点进行拟合,改进标准学习机,改善学习速度和预测性能,最后通过网络流量数据的预测实验验证其可行性。验证结果表明:与其它网络流量预测模型相比,改进极限学习的网络流量预测结果更加可靠,对网络流量将来变化趋势可以更加准确描述,提高了网络流量预测精度。In order to obtain more accurate prediction of network traffic and reduce the congestionfrequency of network, a novel network traffic prediction model based on improved extreme learning machine is proposed in this paper. Firstly, the delay time and embedding dimension are determinedaccording to the chaos of network traffic, and secondly, extreme learning machine is used to simulate the change rule of network traffic which standard learning machine is improved to improve the learning speed and performance, finally, the feasibility of is verified by the network traffic data. The results show, the network traffic prediction results of the proposed model are more reliable Compared with other network traffic prediction models, can describe the change trend of network traffic and improves the prediction accuracy of network traffic.

关 键 词：网络流量 相空间重构 极限学习机 混沌变化特性 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]