摘要
针对传统属性相似度告警聚类不能充分挖掘告警属性语义信息的不足,结合自动交换光网络(automatically switched optical network,ASON)的告警特点,分析了告警属性语义信息对告警聚类的作用,定义了具有分层特点的属性相似度函数,增强了聚类结果的客观性,并利用历史告警库复杂度和聚类扩充率对聚类效果进行了比较分析。实验表明,新方法与传统的告警属性相似度聚类方法相比,具有更高的准确性。对于在特定时间段内故障类型较多的情况,新方法具有较好的适应性。
Aiming at the shortage that the traditional attribute similarity alarm clustering cannot fully exploit alarm attribute semantic information,combining with ASON alarm features,the significance of alarm attribute semantic information for alarm clustering is analyzed,and an attribute similarity function with hierarchical feature is defined,which enhances the objectivity of the clustering results. The clustering results were compared by alarm history library complexity and clustering expansion rate. Experiments showed that,compared with the traditional method,the accuracy of the method in this paper was higher. For a certain period with more types of faults generating,method in this paper was of good adaptability.
出处
《科学技术与工程》
北大核心
2015年第6期210-214,225,共6页
Science Technology and Engineering
基金
中央高校基本科研业务费专项资金(13XS32)资助
关键词
分层属性相似度
聚类
告警关联
ASON
hierarchical attribute similarity
clustering
alarm correlation
ASON
