摘要： 为解决越来越严重的网络拥塞问题,规避恶意网络攻击,提出基于Morlet小波支持向量回归(MWSVR)和自回归移动平均(ARMA)混合模型的流量预测方法。针对较短时间尺度的网络流量,将Morlet小波构造为支持向量回归方法的核,得到MWSVR模型。流量经小波分解成近似和细 节2个部分,使用混合模型中的MWSVR和传统的线性模型ARMA分别预测网络流量的近似部分和细节部分,合成各分量值作为预测结果。与单一的ARMA与MWSVR模型预测结果分别进行比较,结果证明,该混合模型能够较准确地拟合网络中的流量。

关键词: 流量预测, Morlet小波, 支持向量回归, 自回归移动平均, 均方误差

Abstract: In order to solve the increasingly serious network congestion,avoiding malicious attacks,the hybrid model of Morlet Wavelet Support Vector Regression(MWSVR) and Auto Regressive Moving Average(ARMA) are proposed in the thesis.The kernal function of Support Vector Regression(SVR) is replaced by Morlet wavelet function in prediction for short time scales of network traffic which is divided into two parts including approximation and detail with wavelet decomposition.MWSVR model is used to predict the approximate part in the flow,while ARMA model is used to predict the detailed part.Two parts are compounded as the prediction result.Simulation of prediction result shows that the hybrid model predicts the traffic flow in network more precisely than single ARMA or MWSVR model.

Key words: flow prediction, Morlet wavelet, Support Vector Regression(SVR), Auto Regressive and Moving Average(ARMA), Mean Square Error(MSE)

中图分类号: 

• TP393