基于信息扩散原理的关联规则自动评价方法研究  

Research on auto-evaluation of association rules based on information spreading theorem

作　　者：杨炳儒[1] 宋泽锋[1] 侯伟[1] 王映龙[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《计算机应用研究》2008年第10期2965-2969,共5页Application Research of Computers

基　　金：国家自然科学基金重点资助项目(69835001)

摘　　要：在关联规则的评价中,对评价参数的度量是参数在数据库上的平均分布,无法刻画参数在数据库的局部分布;另一方面,评价参数也是有限的。为解决上述问题,首先给出评价关联规则的五个参数及参数选取的原因分析;然后给出分库方案及其分库上的挖掘算法,通过分库挖掘算法获取各参数在分库上的参数值。对获取的参数值,利用扩散原理计算参数在数据库上的分布,通过分布实现对规则的评价和取舍,从而实现关联规则的自动评价和取舍,以及最大程度的自动知识获取,并具有实时查询特性。Regarding evaluation of association rules, existing measurements are based on even distribution of parameters in database, which can not describe the local situation of parameters. Furthermore, few parameters are used in traditional methods. This paper proposed five parameters for evaluating association rules as well as analysis of them at first. Then proposed partition schemes of database as well as corresponding mining algorithm. Values of parameters in different sub-database can be got by using the mining algorithm based on partitioned database. Finally, got distribution of different parameters in sub-database by using information spreading theorem. These kinds of distribution can be helpful to evaluation and selection of rules. Thus auto-evaluation, selection and real-time query are realized to certain extent.

关 键 词：关联规则 信息扩散原理 参数演化 知识发现 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]