针对已有算法复杂度较高,不适用于规模较大网络的问题,将被管系统建立成动态贝叶斯模型,并提出一种能处理多故障的近似推理算法.通过复杂度分析,证明该近似算法时间复杂度为多项式级,远远低于精确算法的时间复杂度下限,可以用于解决大规模动态网络的故障诊断问题.实验结果证明,新算法在准确度方面虽然略低于精确算法,但执行效率上远远高于精确算法.
Fault management is one of the most important parts of network management. It is a challenge problem to quickly and accurately locate the faults of network. Bayesian networks model is a prominent way to solve that problem, but it is limited when the state of the nodes changes over time. Present algorithms based on Bayesian networks model may solve the problem with higher accuracy, but the algorithms are very compex and not proper for large scale of network. A new efficient inference algorithm that can diagnosis multi-fault in dynamic Bayesian networks is proposed. Then, by analysis of its complexity, it is proven that the approximation algorithm has a much lower time complexity than the lower bound of exact algotithm in the dynamic Bayesian networks. Finally, the experiments show that the accuracy of the new algorithm is slightly lower than the exact inference algorithm, but its efficiency is much higher than the exact inference algorithm. This new algorithm can be applied to the large communication networks.