动态指数平滑模型在网络流量预测中的研究

Research on Network Traffic Prediction-based Dynamic Exponential Smoothing Model

网络流量预测是网络性能管理的一个重要组成部分,一种好的预测模型能比较准确地判断网络流量的发展趋势,对网络管理起到推进作用。提出了将最速下降迭代法应用于指数平滑模型的方法,以预测误差平方和(SSE)最小为目标,构造并自动生成了最佳平滑参数,使网络流量的预测模型得以优化,增强了指数平滑模型对时间序列的适应能力,较好地解决了指数平滑预测模型中,平滑参数靠检验确定且为静态,平滑初值难以确定并导致预测偏差等问题。通过分析,证明了此模型能够较准确地预测出网络的流量,使网络管理人员对网络的性能状态有整体的把握并且能够较好地对网络性能进行监测控制,从而提高了网络的服务质量。

Prediction of the network traffic is an important part of the network performance management. This paper proposes a method of applying the steepest-descent iterative algorithm to the exponential smoothing model to predict the smallest of square sum of error （SSE） and construct and generate optimal smoothing parameter automatically to optimize the prediction model of the network traffic and enhance adaptability of exponential smoothing model on time sequence, which solves the problems that smoothing parameter is static and determined only by experience and the smoothing initial value is difficult to determine and to generate prediction deviation and so on. Experiment shows that this model can accurately predict the network traffic, make the administrative operator of the network monitor and control the network performance for the purposes of improvement of network service quality.