基于扩展概念格的属性归纳算法  

Attribute-Oriented Algorithm Based on Extended Concept Lattice

作　　者：王德兴[1,2] 胡学钢[2] 刘晓平[2] 黄冬梅[1] 

机构地区：[1]上海海洋大学信息学院,上海201306 [2]合肥工业大学计算机与信息学院,合肥230009

出　　处：《上海交通大学学报》2009年第3期476-479,484,共5页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金项目(60673028);海外及港澳学者合作研究基金项目(60828005);国家高技术研究发展计划(863)项目(2006AA102239-1);上海市教育委员会科研创新基金(08YZ120);上海海洋大学博士科研启动基金(科07-289)

摘　　要：由于面向属性归纳在进行属性归纳时难以设置适当的阈值,往往会造成概念提升的结果过于一般化或特殊,为此提出了基于扩展概念格的属性归纳算法.在扩展概念格中进行概念提升实现数据泛化,并根据相应的泛化路径,在相应哈斯图上找到合适的泛化阈值,进而得到较好的属性归纳结果.研究结果表明,基于扩展概念格的属性归纳算法在进行属性归纳过程中保留了完备的信息,减少了属性归纳的计算工作量,并具有直观、简捷的特点.When data is generalized or concept tree ascends in the attribute-oriented induction （AOI） the generalization threshold can not have been specified properly, and then the obtained results of concept tree ascension is usually overly generalized or special. The attribute-oriented algorithm based on extended concept lattice was presented, which can perform concept ascension to generalize data, find out the proper generalization thresholds in the Hasse diagram of extended concept lattice according to the corresponding generalization path, then the better results are obtained. The experiments show that the latter keeps the complete information in the attribute-oriented induction, the computing workload is reduced, and has the visual and simple characteristics.

关 键 词：属性归纳 概念格 概念层次 数据挖掘 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]