基于FARIMA-GARCH模型的网络业务预测算法

发布时间：2018-03-01 06:00

本文关键词： FARIMA GARCH CUSUM 流量预测　出处：《通信学报》2013年03期 　论文类型：期刊论文

【摘要】：针对网络流量的波动性与自相似特性为其精确预测提出的挑战,提出了一种基于FARIMA-GARCH模型的预测算法。该算法首先利用分段双向CUSUM检测算法对流量序列的均值进行有效检测,并在此基础上将序列零均值化;然后采用限定搜索法对分数差分阶数进行精确估计;在获得模型参数后,使用FARIMA-GARCH模型对网络流量进行预测。仿真实验表明,限定搜索法能够获得比传统算法更高的估计精度。随后采用真实网络流量验证了预测算法的性能,在保持与FARIMA预测算法等价的时间复杂度下,其均方根和相对均方根误差与RBF神经网络预测算法相当,而高于FARIMA预测算法。同时,预测算法对突发流量的跟踪和预测性能明显优于对比算法,且有更好的区间估计性能。
[Abstract]:In view of the challenges posed by the volatility and self-similarity of network traffic, a prediction algorithm based on FARIMA-GARCH model is proposed. Firstly, the segmented bi-directional CUSUM detection algorithm is used to detect the mean value of traffic sequence effectively. Then the finite search method is used to estimate the fractional difference order accurately. After obtaining the model parameters, the FARIMA-GARCH model is used to predict the network traffic. The simulation results show that, The limited search method can obtain higher estimation accuracy than the traditional algorithm. Then, the performance of the prediction algorithm is verified by using real network traffic, while maintaining the time complexity equivalent to the FARIMA prediction algorithm. The root-mean-square error and the relative root-mean-square error are similar to the RBF neural network prediction algorithm, but higher than the FARIMA prediction algorithm. At the same time, the prediction algorithm can track and predict the burst traffic obviously better than the contrast algorithm, and has better performance of interval estimation.
【作者单位】：电子信息控制重点实验室;电子科技大学通信抗干扰技术国家级重点实验室;
【基金】：国家自然科学基金资助项目(61271168,61001085) 国家科技重大专项基金资助项目(2010ZX03005-002) 国家重点基础研究发展计划(“973”计划)基金资助项目(2009CB320405) The National Natural Science Foundation of China(61271168,61001085)~~
【分类号】：TP393.06

【参考文献】