【作者】 武艺全；

【导师】 丁振国；

【作者基本信息】 西安电子科技大学 ， 计算机系统结构， 2010， 硕士

【摘要】 随着网络规模不断扩大,网络复杂性不断增加,网络故障问题越来越突出。本文针对传统故障诊断中存在的问题,深入地研究了关联规则挖掘与分类挖掘两种数据挖掘方法,并应用于故障诊断中,实现了对网络故障的智能诊断。首先在基于数据挖掘的网络故障诊断定位模型基础上,考虑了网络故障告警信息的群集特点,对基于频繁模式树(FP-tree)的关联规则挖掘算法进行改进,提出了基于频繁模式树的故障群集的诊断定位挖掘FP-treeC算法；根据网络故障告警信息增加的特点,提出了基于频繁模式树的告警数据库中数据增加时关联规则定时增量挖掘FP-treeCT算法；对ID3的决策树挖掘算法进行了改进,实验表明三个算法都在原算法上得到了效率的提高；最后设计了针对某些特定类型的网络故障的一种基于策略的网络故障修复模型PBFR。更多还原

【Abstract】 As the networks have been expanding ever-increasing network complexity, network failures become increasingly acute. Aiming at analyzing the problems in traditional fault diagnosis, this paper presents two kinds of in-depth data mining methods:the association rule mining and classification mining. And they were applied in fault diagnosis to implement intelligent diagnosis of network faults.In the basis of network fault diagnosis which is based on location model-based data mining, together with considering the network fault alarm cluster features, the association rule mining based frequent pattern tree (FP-tree) algorithm is improved, and FP-treeC mining algorithm is proposed to improve the fault diagnosis and location for cluster; according to the network fault features that alarm information increases continuously, association rules incremental update problem about fault diagnosis and location are studied, and an improved timer FP-treeCT incremental updating algorithm for mining association rules is proposed; based on the existing ID3 decision tree mining algorithms, the IID3 algorithm is applied to network fault diagnosis and location. The paper also proposes a plan-based fault restoration model for certain specific types of network fault.更多还原