贝叶斯MLP神经网络在IP承载网性能预测的研究  

Research on performance prediction of IP network by Bayesian MLP neural network

作　　者：侯阿临[1] 李飞[1] 

机构地区：[1]长春工业大学计算机科学与工程学院,长春130012

出　　处：《软件》2013年第4期96-97,127,共3页Software

基　　金：教育部"春晖计划"国际合作科研项目(Z2011138);吉林省科技厅自然科学基金项目(20101523);吉林省科技厅科技支撑计划项目(20100368).

摘　　要：通过对贝叶斯MLP神经网络的学习,发现它在处理大量的实时数据时,具有较高的精准度还能够保证IP承载网的强泛化能力,并且预测结果容易达到运营商的预期。为此尝试将贝叶斯MLP神经网络引入目前IP承载网性能预测,并且通过实验验证贝叶斯MLP神经网络在IP承载网中进行性能预测的可行性结论。The method of Bayesian MLP neural network not only has higher precision in the processing of large amounts of real- time data, but also ensure the generalization ability of IP carrying network, and it is easier to achieve the operator＇s expected from the prediction results. In the paper, Bayesian MLP neural network is introduced into the IP network for performance prediction. The results of simulation show the algorithm of Bayesian MLP neural network can predict the performance of IP network and evaluate its effectiveness.

关 键 词：GPRS漫游信控 支撑网网状网 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]