利用扩展概念格进行关联分类的算法  

A new association rule classification algorithm using extended concept lattice

作　　者：翟悦[1] 郭文书[1] 王立娟[1] 

机构地区：[1]大连科技学院信息科学系,辽宁大连116052

出　　处：《辽宁工程技术大学学报（自然科学版）》2015年第11期1280-1284,共5页Journal of Liaoning Technical University (Natural Science)

摘　　要：针对关联分类规则产生的候选规则过多导致效率不高的问题,提出一种基于频繁闭项集组成的扩展概念格的分类规则获取方法.利用频繁闭项集提出一种新的概念格模型,通过性质和定理对概念格结点进行剪枝,以抽取分类尽量少且最有效的关联分类规则.研究结果表明:该算法能挖掘出高质量且包含重要信息的关联分类规则,并大大减少关联分类规则的数量,在分类准确率上比现有的关联分类典型算法更高.Associative classification rules generate candidate rules too much for a given dataset. This paper proposed a novel classification rule acquisition method based on the extended-lattice that is composed of the frequent item sets. Firstly, a new and more advantageous lattice structure based on frequent closed itemsets is proposed. Secondly, theorems and properties of the lattice were used to prune the branches of the lattice and reduce redundant rules quickly. The experiment shows that the proposed method is capable of extracting highly useful rules representing key information of the data and in the same time reducing the number of rules significantly. Compared with the typical associative classification algorithms, the proposed method can mine much less number of rules and has higher average classification accuracy.

关 键 词：数据挖掘 频繁闭项集 扩展概念格 关联分类规则 规则剪枝 

分 类 号：TP301.06[自动化与计算机技术—计算机系统结构]