节点文献

基于序列模式挖掘的故障管理系统设计与实现

Design and Implementation of Fault Management System Based on Mining Sequential Patterns

【作者】 郭文婷

【导师】 石坚;

【作者基本信息】 华中科技大学 , 电磁场与微波技术, 2007, 硕士

【摘要】 随着通信和信息技术的飞速发展,现代网络的规模越来越大,网络复杂性日益提高,设备的分散性、接口的多样性以及各设备厂家之间设备的相对独立性,这一切使得建设具备跨平台性、高效等特性的网络故障管理系统的需求越来越明显。本文参考国内外相关领域最新研究成果,在对现有网络故障管理系统分析的基础上,综合考虑了网络故障管理的任务、告警数据的特点等因素,利用数据挖掘技术中的序列模式挖掘来实现告警相关性分析,制定了较为合理的故障管理业务处理流程,在整体上实现了对网络资源的故障管理功能。本文主要做了如下工作:(1)根据故障管理系统的功能需求,设计故障管理的业务处理流程、系统体系结构以及数据库结构;(2)从原理上研究如何建立告警相关性分析模型,着重分析如何针对告警数据进行数据挖掘工作,提出一种新的序列模式挖掘算法,并且基于该算法的框架提出一种增量式告警关联规则挖掘算法,解决了挖掘参数不变、告警数据增加导致的重复挖掘情况,提高数据挖掘的效率;(3)提出拓扑约束的方法,利用网络中设备之间的拓扑连接知识来对告警关联规则挖掘工作进行约束,减少同步时钟失调情况下挖掘结果的误差;(4)对系统所采用的序列模式挖掘算法的性能与准确性进行测试,并对系统的各功能进行测试。

【Abstract】 With the rapid development of the technology of communication and information, nowadays, the internet has increased fleetly in scale and complexity. The devices have the characteristic of decentralization, the interfaces are diversiform, and the devices from different manufactory are independent comparatively, all above make the demand of building a multi-platform and efficient network fault management system become more and more heavy.This paper is achieved after consulting the latest pertinent research production and analysing the situation of the network fault management system. After considering the task of network fault management and the characteristic of alarm data synthetically, it completes alarm correlation analysis using mining sequential patterns of the data mining technology, establishes comparatively reasonable processing course of the operations in fault management, and actualize the function of fault management for network resources. In this paper, we innovate in the following points:(1)Design processing course of the operations, the framework of the system, and the structure of the database, considering the function requirement of the system.(2)Do research about how to establish the module of alarm correlation analysis, and analyse the method of data mining upon alarm data. Put forward a new mining arithmetic, and then bring forward an incremental mining arithmetic of alarm correlated rules. It resolves the repeated mining situation when the mining parameter is unaltered and alarm data is increased, and the efficiency of data mining is enhanced.(3)Bring forward the method of topology restriction, restrict the correlated rules mining using the topology connection knowledge between the devices in the network. It can decrease the error of mining result under the circumstance synchoronization clock maladjustment.(4)Make test to prove the performance and precise of the arithmetic adopted by the system, and test the functions of the system.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络