一种基于神经网络的网络流量组合预测模型  

Combined prediction model of Internet traffic based on neural network

作　　者：冯海亮[1] 陈涤[1] 林青家[1] 陈春晓[1] 

机构地区：[1]山东大学信息科学与工程学院,山东济南250100

出　　处：《计算机应用》2006年第9期2206-2208,2228,共4页journal of Computer Applications

基　　金：山东省自然科学基金资助项目(Y2001G05);Cisco教育科研资助计划资助项目(PO135006928)

摘　　要：Internet流量是具有相关和非平稳特性的时间序列,文中通过对过去一些流量模型的分析,构建了一种新的预测模型。该模型首先使用小波方法对网络流量进行预处理,然后分别使用线性神经网络和E lman神经网络进行预测,以保证能够描述流量的相关和非平稳性,最后把两种预测的结果通过BP神经网络合成为最终预测结果。通过对不同流量的一步预测和多步预测仿真,验证了组合模型比单一模型具有更高的预测精度。The Internet traffic is correlated and non-stationary time series. By analyzing some past traffic models, a new prediction model was established. In this model, Internet traffic was pretreated with wavelet methodat first. Then, linear NN （Neural Network） and Elman NN were used respectively to make prediction. Therefore, the correlation and non-stationary characteristics of the traffic could be described. Finally, the two predictions were combined into the final result through BP neural network. Through one-step and multi-step prediction simulations on different kinds of traffic respectively, it is verified that the combined model can predict more precisely than the individual model.

关 键 词：流量 相关 非平稳 神经网络 预测 

分 类 号：TN915.01[电子电信—通信与信息系统]