【作者】 陈振伟；

【导师】 郭拯危；

【作者基本信息】 河南大学 ， 应用数学， 2008， 硕士

【摘要】 实现网络QoS控制对于网络管理及维护相当重要,而网络流量预测对于实现网络的QoS控制可以起到十分重要的作用;另外,流量预测在入侵检测中的应用也受到越来越多的关注。网络流量预测极具现实研究意义。网络流量的预测要求较高的实时性,另外准确率也是一个很重要的指标。本文基于神经网络预测模型,并引入小波函数对其进行改进,建立了一个新的网络流量预测模型;另外针对几种神经网络的不同特点进行有效的整合,建立了一个新的网络流量预测模型;并对以上两种模型写出了相应的算法,仿真试验证明,该模型预测误差低,并且具有普适性。本文研究的内容是对神经网络流量预测模型进行改进,并设计相应的预测步骤,主要有以下研究成果:1、在BP神经网络预测模型的隐含层引入小波函数作为其传递函数,建立一个新的小波神经网络预测模型,并设计了相关的预测算法,提取了真实的网络流量对该模型进行了仿真验证,在保证预测精度的前提下,提高了神经网络的训练速度,达到了改进的目标。2、针对线性神经网络,Elman神经网络,RBF神经网络,BP神经网络等不同预测模型优缺点,对它们进行有效的整合,利用这些神经网络的优点,建立了一个基于多神经网络的网络流量预测模型,并用真实的网络流量对该模型进行仿真验证,提高了其预测性能。更多还原

【Abstract】 Realizing the network Qos control for network management is of the same importantance as the network maintence,and network traffic prediction is very important to realization of the network Qos control;Another, the application of network traffic prediction in intrstion detection system Caused concern more and more. network traffic prediction is quite a useful research which have essential reality meanings.The request for the quality of real-time is more than other aspects for the estimation of the network discharge .Except for that,accutacy is also an important sign .This text is based on the model of the neural network prediction .To improve the basic function ,we import the wavelet function to create a new network traffic prediction model .Also we establish new valid integration prediction model based on the different characters of several neural-networks.I give the correspond algorithm of the above two models , Simulation experiment certificate that the model has lower error,and the model can be used generally.The content of this text is aimed at the improvement of the neural network traffic prediction model, and the design of the correspond estimating step, main research results are:1.I import the wavelet function as the passing function of the hidden layer in the BP neural network traffic prediction model,create a new wavelet neural network traffic prediction model,design the related prediction algorithm, and withdraw the true network traffic to verificate the model.In this way ,we improve the training speed of the neural network and the target of improvement comes true.2.We made valid integration based on the advantages of the model: the line neural network , Elman neural network, RBF neural network ,and BP neural network .We make use of the advantages of the above several of neraul network models and get satisfacting result by building a model based on several of neraul networks and doing the imitating verification for this new network traffic prediction model using the true network traffic.更多还原