基于神经网络和遗传算法的VBR通信流量预测  

VBR Traffic Forecasting Based on Neural Networks and Genetic Al gorithm

作　　者：潘昊[1] 吴尚[1] 钟珞[1] 

机构地区：[1]武汉理工大学计算机学院,武汉430070

出　　处：《计算机工程与应用》2004年第6期11-13,17,共4页Computer Engineering and Applications

基　　金：国家"十五"重点科技攻关项目(编号:2001BA307B01-0201)

摘　　要：论文阐述了利用神经网络预测由连续自动回归(AR)马尔可夫模型所代表的可变位速率通信流量(VBR);在这一理论的基础上,介绍一个BP神经网络模型,它采用拆分,组装方法来构造一个学习结果达到均方根误差全局最小点的BP神经网络,该方法结合遗传算法能有效克服局部极小点,缩短学习时间和减小学习难度,并能够精确地预测VBR通信流量。Forecasting VBR traffic which represented by continuous automatic return Markrov Model by neural networks.Based on this theory,a BP neural networks model is introduced which constructs a study result to achieve a minimum point of RMS error though the method of split and assembling.Combining this method width genetic algorithm can con-quer local minimum point ,shorten study time and reduce the study difficulty and forecast VBR traffic precisely.

关 键 词：神经网络 拆分与组装 通信流量预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]