属性拓扑关联规则发现

Association Rules Detecting Based on Attribute Topology

在形式概念分析领域,属性拓扑理论提供了简洁,明确,可视化的概念计算的方法.然而,概念要求属性与对象间形成充要的双向映射关系,这一要求在大规模数据挖掘下往往过于严格.因此,本文以属性拓扑为基础,基于支持度和置信度的本质要求,提出一种属性拓扑关联规则发现算法,该方法首先构建频繁净化形式背景,由属性拓扑,直接发现二元频繁模式,并由BFSW子算法,计算三元及以上频繁模式,经过置信度检验,进而获得所需的关联规则.该算法弥补了传统概念计算中忽略属性间关联规则知识发现的不足,提供了发现属性对象间充分不必要关系的有效视角.

In FCA, Attribute Topology provided several methods to compute formal concept. The process of AT＇s computing is simple, clear and visual. However,concept demands the necessary and sufficient bidirectional mapping between attributes and objects. This re- quirement is so strict especially in large scale data mining. Therefore, this paper proposed an association rules detecting algorithm based on AT and the nature requirements of Support and Confidence. UsingAT constructed by frequent pure context,this algorithm can get binary frequent patterns immediately, and compute the other patterns by BFSW algorithm. After that, the association rules can be computed by confidence test. This algorithm made up the drawback of losing sight of associations among attributes in traditional concept computing,provided a method to discover the unnecessary and sufficient mapping between attributes and objects.