ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2015, Vol. 52 ›› Issue (5): 1137-1145.doi: 10.7544/issn1000-1239.2015.20131757

• 网络技术 • 上一篇    下一篇

遗传算法优化回声状态网络的网络流量预测

田中大1,高宪文2,李树江1,王艳红1   

  1. 1(沈阳工业大学信息科学与工程学院 沈阳 110870); 2(东北大学信息科学与工程学院 沈阳 110819) (tianzhongda@126.com)
  • 出版日期: 2015-05-01
  • 基金资助: 
    基金项目:国家自然科学基金重点项目(61034005);辽宁省博士科研启动基金项目(20141070)

Prediction Method for Network Traffic Based on Genetic Algorithm Optimized Echo State Network

Tian Zhongda1, Gao Xianwen2, Li Shujiang1, Wang Yanhong1   

  1. 1(College of Information Science and Engineering, Shenyang University of Technology, Shenyang 110870); 2(College of Information Science and Engineering, Northeastern University, Shenyang 110819)
  • Online: 2015-05-01

摘要: 网络流量预测是网络拥塞控制与网络管理的一个重要问题. 网络流量时间序列具有时变、非线性特征,导致传统时间序列预测方法预测精度比较低,无法建立精确的预测模型.回声状态网络(echo state network, ESN)在非线性混沌系统预测与建模方面有着良好的性能,非常适合网络流量的预测.为了提高网络流量的预测精度,提出一种基于遗传算法(genetic algorithm, GA)优化回声状态网络的网络流量非线性预测方法.首先利用回声状态网络对网络流量进行预测;然后利用遗传算法对回声状态网络预测模型中的储备池参数进行优化,提高预测模型的预测精度.通过中国联合网络通信公司辽宁分公司采集的实际网络流量数据进行了仿真验证.与差分自回归滑动平均模型(auto regressive integrated moving average, ARIMA)、Elman神经网络以及最小二乘支持向量机(least square support vector machine, LSSVM)这3种常见预测模型进行了对比,仿真结果表明提出的方法具有更高的预测精度与更小的预测误差,更能刻画网络流量复杂的变化特点.

关键词: 网络流量, 非线性, 预测, 遗传算法, 回声状态网络

Abstract: Network traffic prediction is an important problem of network congestion control and network management. Network traffic time series have time-varying and nonlinear characteristics, So the prediction accuracy of traditional time series prediction method is relatively low and it is unable to establish accurate prediction model. Echo state network (ESN) has good performance in prediction and modeling of nonlinear chaotic system and is very suitable for network traffic time series prediction problem. In order to improve the prediction accuracy of network traffic, the network traffic nonlinear prediction method based on genetic algorithm optimized echo state network is proposed. Firstly, echo state network is used for network traffic predicton, then the genetic algorithm is used to optimize the parameters of the reservoir of the echo state network prediction model. Finally, the prediction accuracy of the prediction model is improved. The actual network traffic data from China Unicom in Liaoning Province is used for simulation verification. Three common prediction models including auto regressive integrated moving average (ARIMA) prediction model, Elman neural network prediction model and least square support vector machine (LSSVM) prediction model are compared, and the simulation results show that the proposed method has higher prediction accuracy with smaller prediction error and it can describe the network traffic characteristics of complex changes.

Key words: network traffic, nonlinear, prediction, genetic algorithm(GA), echo state network(ESN)

中图分类号: