摘要
传统通信网络告警处理方法主要由维护专家依据经验判断形成处理规则并固化在网络告警系统中进行实现,然而该人工维护方式难以适应海量数据环境下实时通信告警规则的处理需求。为此,提出一种基于加权频繁模式树(WFP-tree)算法的告警规则自动挖掘方法,将原始告警数据按时间窗口方式进行分段处理,通过BP神经网络、支持向量机、层次分析法生成告警设备的权重信息,并采用WFP-tree算法自动挖掘加权频繁项集。实验结果表明,与传统Apriori和FP-growth算法相比,WFP-tree算法在通信网络告警分析方面具有更好的频繁项压缩效果及更强的重要关联规则发现能力。
Traditional communication network alarm correlation rules are often manually done by experts and coded into network fault management systems. However, the artificial maintenance method is difficult to meet the huge amounts of data processing requirements of real-time communication alarm rules. To solve this problem, this paper proposes an automatic alarm rule mining method based on Weighted Frequent Pattern-tree(WFP-tree) algorithm. It uses the sliding window method to convert raw data into alarm transactions, and employs BP neural network, Support Vector Machine (SVM) and Analytic Hierarchy Process (AHP) methods to generate the weight information of alarm equipment. Finally, it uses WFP-tree algorithm to automatically generate the weighted frequent itemset. The experimental results show that, the WFP-tree algorithm performs better in frequent itemset compression and important domain correlation rule finding compared with Apriori and FP-growth algorithms.
出处
《计算机工程》
CAS
CSCD
北大核心
2016年第4期190-196,共7页
Computer Engineering
基金
上海市科技创新行动计划基金资助项目(13511505200)
上海市科技人才计划基金资助项目(14XD1421000)
上海财经大学2014年研究生创新基金资助项目(CXJJ-2014-438)
关键词
通信网络告警
关联规则
权重因子
加权频繁项集
FP-GROWTH算法
加权频繁模式树算法
支持度
communication network alarm
correlation rule
weighted factor
weighted frequent itemset
FP-growth algorithm
Weighted Frequent Pattern-tree (WFP-tree) algorithm
support degree