本文通过分析移动通信网络中无线、交换、传输故障告警数据的分布模型和特点，提出了一种基于BP网络的多级告警筛选模型。该模型在实时采集移动通信网络中告警数据的基础上，通过对告警数据的过滤和预处理将告警数据归一化，并使用二进制编码机制将告警数据编码为可使用神经网络进行处理的类型，进而通过使用高斯感知器对告警数据进行预分类，有效降低BP网络需要处理的故障告警数据量，最后通过利用BP网络的自适应特点和模式识别能力，对告警进行模式识别分类，定位主要故障点，提高了移动网络告警的监控效率和网络的稳定性。

A novel multilevel fault alarming filter model is presented by studying the distributions and characteristics of the fault alarms in the radio, switching and transferring of mobile communication networks. Firstly, the fault alarms, which are collected in a realtime way, are filtered and preprocessed to be normalized. Secondly, the fault alarms are coded in binary, and the result data are used by the Gaussian perception network to preclassify them to reduce the alarms that the BP network are supposed to process. Finally, the points of failure are located by the adaptive ability and pattern recognition capability of the BP network. The total system can ensure the network to be stable, efficient, and reliable.