可拓关联规则在告警相关性分析中的研究与应用

发布时间：2018-05-08 14:35

本文选题：数据挖掘 + 关联规则　；参考：《沈阳理工大学》2017年硕士论文

【摘要】：网络缩短了世界的距离,作用日益重要,网络故障的影响也越来越严重。传统的网络故障诊断方法主要依靠工作人员在长期处理故障的过程中所积累的经验,这种方法效率较低且培养一个有经验的技术人员花费的成本较高、培养周期长。随着数据挖掘技术的逐渐成熟及在社会生活中的成功应用,将其应用于网络故障告警相关性分析也成为研究的热点。分析故障告警之间相互关联的规则并应用于网络故障定位,能够提高故障定位的效率,具有很高的实际应用价值。本课题以数据挖掘技术和可拓学知识为基础,以通信网络故障告警相关性分析及故障点定位为目的,完成可拓关联规则应用于告警相关性分析的理论算法研究,并通过合成数据集,运用C#语言进行编程做仿真实验,通过Matlab软件作图,以图形化的形式进行数据分析,验证算法的合理性。首先,详细介绍基于Apriori算法的网络故障告警相关性分析实例,分析此算法及其改进算法在告警相关性分析中存在的瓶颈。然后,在此基础上根据网络故障告警数据的特点提出本课题的算法。根据可拓学的基元理论以及告警属性的特征,采用可拓物元模型对告警信息中的关键属性进行形式化表达,使其适用于数据挖掘算法并且从根源上减少数据量;在添加水平权值的基础上,根据告警信息的时间特性,添加垂直权值,这种混合加权算法可以减少冗余规则的产生,提高故障诊断的效率;在设置水平权重系数的过程中为减少主观因素的影响,根据告警实例属性的特征,考虑各属性值对水平权值的客观作用,运用可拓关联函数的方法计算告警之间的相似度并确定权重值。为减少对告警实例库的扫描次数,在关联规则挖掘过程中,采用集合的方式存储项集,能够有效降低时间复杂度。通过仿真实验,验证算法的合理性。最后,分析网络故障实时诊断的必要性及研究意义。根据可拓关联规则的特点,结合可拓距与可拓变换理论提出新发告警与可拓关联规则的匹配方法。
[Abstract]:Network shortens the distance of the world, plays an increasingly important role, and the impact of network failures is becoming more and more serious. The traditional network fault diagnosis method mainly relies on the experience accumulated by the staff during the long period of dealing with the fault. The efficiency of this method is low and the cost of training an experienced technician is high and the training period is long. With the maturity of data mining technology and its successful application in social life, the application of data mining technology to network fault alarm correlation analysis has become a hot topic. It can improve the efficiency of fault location by analyzing the correlation rules of fault alarm and applying it to network fault location, which has high practical application value. Based on the technology of data mining and extension knowledge, and aiming at the correlation analysis of fault alarm and the location of fault point in communication network, the research on the theoretical algorithm of applying extension association rules to alarm correlation analysis is completed in this paper. The simulation experiment is carried out by using C # language, and the data is analyzed graphically by Matlab software, and the rationality of the algorithm is verified by synthesizing the data set and using C # language to do the simulation experiment. Firstly, an example of network fault alarm correlation analysis based on Apriori algorithm is introduced in detail, and the bottleneck of this algorithm and its improved algorithm in alarm correlation analysis is analyzed. Then, according to the characteristics of network fault alarm data, the algorithm of this topic is proposed. According to the basic theory of extenics and the characteristics of alarm attributes, the key attributes in alarm information are formally expressed by extension matter-element model, which is suitable for data mining algorithms and reduces the data volume. On the basis of adding horizontal weights and adding vertical weights according to the time characteristics of alarm information, this hybrid weighting algorithm can reduce the generation of redundant rules and improve the efficiency of fault diagnosis. In order to reduce the influence of subjective factors in setting the horizontal weight coefficient, the objective effect of each attribute value on the level weight value is considered according to the characteristics of alarm instance attribute. The method of extension correlation function is used to calculate the similarity between alarms and determine the weight value. In order to reduce the number of scans on the alarm case library, in the process of mining association rules, the item set can be stored in the form of set, which can effectively reduce the time complexity. The rationality of the algorithm is verified by simulation experiments. Finally, the necessity and significance of real-time network fault diagnosis are analyzed. According to the characteristics of extension association rules, combined with extension distance and extension transformation theory, the matching method of new alarm and extension association rules is proposed.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TP393.06

【参考文献】