In the communication network, since the interconnection of a large number of components, mobile network operators run Operations Support Systems that generate vast amounts of alarm events. The harsh challenge for network operators is how to find the potential root causes from massive alarms in real time. In this paper, we propose a novel solution for the root causes analysis. The solution includes a silent gap based approach to resolve the asynchrony of alarms, an algorithm for constructing Bayesian network (BN) based on sequentiality between alarms, and Bayesian inference to identify the root causes. The silent gap-based approach reduces preprocessing time while taking into account the validity. Also, the proposed BN-based mechanism allows the identification of the root causes with a higher accuracy. Experiments conducted on a real alarm dataset are provided to support the proposed methods. In addition, we propose a new algorithm processing framework.

在通信网络中，由于大量组件的互连，移动网络运营商运行的运营支持系统会生成大量警报事件。对于网络运营商而言，严峻的挑战是如何从大量警报中实时查找潜在的根本原因。在本文中，我们为根本原因分析提出了一种新颖的解决方案。该解决方案包括解决警报异步问题的基于静默间隙的方法，基于警报之间的顺序构造贝叶斯网络（BN）的算法以及用于确定根本原因的贝叶斯推理。基于沉默间隙的方法在考虑有效性的同时减少了预处理时间。而且，提出的基于BN的机制允许以更高的准确性识别根本原因。提供了对真实警报数据集进行的实验以支持所提出的方法。另外，我们提出了一种新的算法处理框架。