基于序列模式挖掘的网管告警系统

告警关联分析是网络故障管理中的一个难点,传统方法由于需要引入大量的先验知识而难以适应网络复杂多变的情况.序列模式挖掘作为一种时序数据分析的有效手段,能够自动从告警中提取出有助于关联分析的情景规则.本文首先介绍了与情景规则有关的一些基本概念,然后讨论了挖掘情景规则的常用算法,进而提出了基于序列模式挖掘的告警关联分析的网管系统模型,并实现了一个简单的原型系统,实验证明该系统能够从海量告警中提取出有助于关联分析的情景规则.

Alarms Correlation Analysis System Based on Sequential Pattern Mining

Alarms correlation analysis is a difficult problem in network fault management. Traditional methods in this field can hardly work well when network is complex and changeful and the reason is that too much prior knowledge is needed in these methods. As an effective means to analyze timed data sequential pattern mining can extract episode rules from alarms, which is helpful to analyze correlation. In this paper some concepts related to episodes rules are introduced firstly and then the common algorithm for mining episode rules is discussed. Then an alarms correlation analysis system model based on sequential pattern mining is presented. In addition a simple prototype system is implemented in this paper and the experiment indicates that the system can extract episode rules helpful for correlation analysis.