利用贝叶斯网络建立通信网告警相关性模型,采用EM算法对不完全观察的隐变量进行学习.介绍了基于贝叶斯网络的基本概念.提出了通信网功能分层结构的思想,建立不同网络层次间的故障传播模型.讨论了从故障传播模型中构造贝叶斯网络.结合SDH over DWDM实验模型,具体讨论了贝叶斯参数学习的实现步骤及结果.
The paper proposes an alarm correlation model based on Bayesian networks among communication networks. It adopts EM algorithm to learn the hidden variables in Bayesian networks. The basic concepts of Bayesian networks are introduced. Thhe paper presents a hierarchical architecture for large communication networks. The fault propagation model is used to model the functional relationship among the sub-networks. The paper also discusses how to construct Bayesian networks from the fault propagation model. According to SDH over DWDM experimental systems, the realization and results of the Bayesian learning are discussed.