基于小波分析的ARMA-SVR网络流量预测方法

摘要

Aiming at improving the prediction accuracy of network traffic,a prediction model based on wavelet analysis was es-tablished.The traffic signal was divided into several detail ones of high frequency and an approximate one of low frequency,and they were predicated separately using autoregressive moving average (ARMA)and support vector machine (SVR)methods. These prediction results were combined linearly to obtain the final prediction traffic.In this way,the method ensures the fitting accuracy of the approximate signal and avoids over-fitting phenomenon of the detail signals.Compared with other prediction methods,the results of contrastive simulations show that the algorithm can effectively improve the prediction accuracy of net-work traffic.%为提高网络流量预测精度，采用一种基于小波分析理论的预测模型。通过将网络流量分成多个高频的细节信号和一个低频的近似信号之和，分别采用 ARMA模型和 SVR模型对细节信号和近似信号进行预测，将各部分的预测结果进行线性组合，得到最终的预测结果，在确保近似信号拟合精度的同时，避免细节信号的过拟合。将该模型和其它预测模型的预测误差进行仿真对比分析，分析结果表明，该算法能有效改善网络预测模型精度。