一种基于异常点检测的电信网络性能监控策略a-电子与信息学报.pdfVIP

第 31 卷第 9 期 电 子 与 信 息 学 报 Vol.31No.9 2009 年 9 月 Journal of Electronics Information Technology Sept..2009 一种基于异常点检测的电信网络性能监控策略 于艳华 宋俊德 (北京邮电大学计算机科学与技术学院 北京 100876) 摘 要：电信网络运营维护方式在由“面向网络”向“面向客户”转变，电信网络性能管理也需要由被动地监测向 主动地监控转变。主动性能监控机制通过检测代表性能降质的异常点来实现故障的快速恢复。该文提出一种基于支 持向量机时间序列预测(Support Vector Machines, SVM)和相关置信区间的异常点检测机制。另外，提出一种新的 支持向量机时间序列建模的自由参数的选取方法，并给出计算机实现过程。理论分析和实验结果表明了该自由参数 选取方法的正确性和异常点检测机制的有效性。 关键词：电信网络；主动性能监控；时间序列预测；支持向量机；白噪声；置信区间 中图分类号：TN915.07 文献标识码：A 文章编号：1009-5896(2009)09-2220-06 A Mechanism of Telecommunication Network Performance Monitoring Based on Anomaly Detection Yu Yan-hua Song Jun-de (The School of Computer Science and Technology, Beijing University of Posts and Telecommunications, Beijing 1000876, China) Abstract: With operation and maintenance mode of telecommunication network changing from ‘network oriented’ to ‘subscriber oriented’, network performance management should also be changed from passive monitoring to proactive monitoring. Proactive Performance Monitoring (PPM) enables a fast fault correction by detecting anomalies designating performance degradation. In this paper, a novel anomaly detection approach is proposed taking advantage of time series prediction and associated confidence interval based on Support Vector Machines (SVM). In addition a novel meta-parameters selection approach is proposed by checking if the training residual is white noise. Theoretical analysis and experimental results verify the correctness of the meta-parameter selection approach and the effectiveness of anomaly detection mechanism.