基于扩展概念格的带约束关联分类规则挖掘方法  

Constrained Association Rule Classification Approach based on Extended Concept Lattice

作　　者：翟悦[1,2] 李楠[1] 于文武 ZHAI Yue;LI Nan;YU Wenwu(College of Digital Technology,Dalian University of Science and Technology,Dalian 116052,China;Software Institute,Dalian Jiaotong University,Dalian 116028,China)

机构地区：[1]大连科技学院数字技术学院,辽宁大连116052 [2]大连交通大学软件学院,辽宁大连116028

出　　处：《大连交通大学学报》2021年第4期88-93,共6页Journal of Dalian Jiaotong University

摘　　要：提出一种改进的带约束的关联分类规则挖掘方法.采用扩展概念格结构存储数据库中全部的频繁项集以保证信息不丢失,通过相关定理削减大量候选项目集,并引入差集概念加快了对类属性的支持度和置信度计算速度,从而加快格构建过程,同时减少了格结点占用空间,在之前构建完成的扩展概念格上根据给定的约束条件进行规则提取.在三种不同的数据集上对所提出的方法进行挖掘时间与内存使用测试,并与已有关联分类算法进行对比分析,实验结果表明所用方法不仅在时间和空间上性能良好,还满足重用性要求.Class association rules mining often consider the constraints of individual needs. A novel classification rule acquisition method is proposed based on the extended-lattice. In order to avoid excessive candidate rules, lattice structure which stores all frequent itemsets is proposed. It looks up the node which contains the constrained itemsets in the lattice, and the diffsets can accelerate the computation speed and occupy less memory. Three different types of data sets of the proposed method are tested for mining time and memory usage. The experiment result shows that the proposed method has good performance in time and space and also meets the requirement of reuse.

关 键 词：数据挖掘 约束条件 扩展概念格 关联分类规则 差集 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]