【作者】 王仲佳；

【导师】 欧阳继红；

【作者基本信息】 吉林大学 ， 计算机应用技术， 2005， 硕士

【摘要】 基于经典的FP_growth 关联规则,本文提出了具有动态加权特性的改进算法WFP_growth。把事务数据库中的项目按其重要程度划分为5个等级;运用层次分析(AHP)算法构造判断矩阵,计算特征向量;将得到的向量作为权值,与项目在事务数据库中出现次数综合考虑作为衡量重要程度的标准,生成FP_tree;最后得到频繁项目集和关联规则。由于权重的赋予过程可以由领域专家动态改变,不但能挖掘出更有意义的规则,而且在算法的运行初期就大量剔除了那些权重小的无用项目集,从而大大提高了算法的运行效率。本文将WFP_growth 应用于通信网络管理中进行故障辨析和定位,对已有的网络管理软件(OSSManager)采集来的告警信息进行相应的处理后结合滑动窗口技术生成事务数据库,在此数据库上运行WFP_growth 算法得到告警之间的关联规则以及告警设备的关联规则。为突破网络管理系统知识获取瓶颈提供了一个有效的解决方案。更多还原

【Abstract】 Data mining or Knowledge discovery in Database (KDD) is an merging field, which is to extract implicit, previously unknown, and potentially useful information out of large amounts of collected data . Mining association rule is an important content of data mining. Association is the discovery of association relationships or correlations among a set of items. They are always expressed in the rule form showing attribute-value conditions that occur frequently together in a given set of data . Mining association rule is a rule that all the support and confidence are more than minsup and minconf, given by customers, respectively. On the basis of traditional association rule algorithm, this text puts forward an Improved Association Rule Algorithm with the Characteristic of Dynamic Added Weight, then introducing its application in fault management of telecom. under the frame of traditional algorithm, each item is processed in equal and consistent way in the database, usually using the frequent degree (support) to measure their importance. However when the item of database are asymmetric distribution and have bigger discrepancy in the frequency of the appearance, it will cause hard enactment of minsup, if establish high, the association rules will may can not involve to item which appear the lower of frequency; But if establish low, will discover too many unmeaning association rules, may still cause combination explode. The association rule of added weight aim at the process of obtaining the association rule, and carry on the adjustment of this set of item’s support. It mainly considers some circumstances of actuality, making some product which profit high rates but not trading frequent also can match the minimum support. Therefore need to promote the weight of those high value products, make them becoming the frequent item set, so as to find out its related rule. In order to weaken the influence of the subjective factor in the process of added weight, we adopt layer analysis method (AHP) which from a famous strategist of American A. L. Saaty puts forward to certain value of index weight, that method only need to ask expert to give comparison the importance between these two index sign, thus divide item of business database into five level according to its important degree. Make use of AHP to judge matrix, and then to calculate vector of characteristic; The vector that will get as the value of weight to compute heavy of weight of each other. After assigning the value of weight for each item, we make use of support of added weight that had gotten to construct the WFP _tree and to produce the item of frequent combination gradually. Be different from association rules algorithm of added weight under the APRIORI, we need to product frequent combination gradually by electing the candidate, then put the problem of frequent mode with long detection return to some short modes, then use the method of linking suffix to lower the expense for manhunt consumedly, all above just by building up a WFP _ tree. Under the environment of experiment that we built to make a detail comparison between the WFP_growth and the other traditional algorithm of association rule, we find that the conclusion of experiment and analysis of theories all show that this algorithm has a good characteristic of time and space. At the last part of this article, we combine the actual network environment of the telecommunication in jltele, putting forward model of warning sequence mode exhumation of the trouble of telecommunication. That model homologous carries on warning information which the network management software (OSSManager) had already collected , then combining to technique of glide window way to produce the business database , after that to circulate the algorithm of WFP_growth under this更多还原