基于分数差分和Fuzzy-AR的网络流量建模和预测  

Network Traffic Modeling and Prediction Based on Fractional Differencing and Fuzzy-AR

作　　者：胡申敏[1] 许维胜[1] 王中杰[1] 余有灵[1] 

机构地区：[1]同济大学电子与信息工程学院,上海200092

出　　处：《计算机工程与应用》2006年第19期104-107,共4页Computer Engineering and Applications

基　　金：国家863高技术研究发展计划资助项目(编号:2003AA412310);上海市高校优秀青年教师培养计划(2004)

摘　　要：文章提出了一种利用分数差分和Fuzzy-AR(模糊自回归模型)进行网络流量建模和预测的新方法。这种方法既能刻画实际网络流量的长相关性,又能描述其中的非平稳和非线性分量,同时具有较低的辨识复杂度。这个方法的两个部分建模和预测是密切相关的。首先它们都通过分数差分的方法消除时间序列中的长相关性,然后分别用模糊自回归模型进行建模或预测。实验表明相比传统的模型,这种方法的预测更加有效。This paper proposes a new method to model and predict network traffic based on fractional differencing and fuzzy-AR model.The new method can capture not only the long-range behavior of network traffic but also the nonstable and non-lineal components of it.Moreover,this new method alleviates the complexity of the estimation of model parameters.This new method includes two relative parts：the modeling algorithm and the prediction algorithm.Both of them use the technique of fractional differencing to remove the long-range dependence in the time series.Then they model or predict the time series respectively by fuzzy-AR model.Our experiments of real network traffic illustrate that the prediction results using our method are better than those using the traditional models like AR.

关 键 词：长相关 自相似 模糊自回归模型 模糊C-自回归模型聚类 分数差分 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]