一种启发式属性约简算法  

A Heuristic Algorithm of Attribute Reduction

作　　者：丁守祯[1] 桑琳[1] 宫悦[1] 肖厚国[1] 

机构地区：[1]大连海事大学数学系,辽宁大连116026

出　　处：《微计算机信息》2008年第18期230-232,共3页Control & Automation

基　　金：国家自然科学基金项目(50575026);辽宁省教育厅资助项目(2005063)

摘　　要：本文在研究前人算法的基础上,采用分而治之的思想。考虑信息分布区域的不同对知识划分的影响。提出了两种改进的属性重要度定义;针对文献中加权平均属性重要度中,加权参数人为确定的局限性,对权数进行了改进。根据粗集的拓扑特征,构造了相应的启发式信息,最后,通过实例证明了算法的有效性。On the basis of research of previous attribute reduction algorithms. A modified heuristic algorithm of attribute reduction is presented in this paper. This algorithm adopts an idea of divide and rule. Two improved definitions of attribute significance are proposed considering the influence of information area of distributing to knowledge compartmentalization. The shortage of predefined weight of condition attribute in calculating weighted sum attribute significance in is pointed out and the weight is modified at the same time. With a view of characteristics of analysissitus, corresponding heuristic information is constructed. Finally, the validity and feasibility of the algorithm is demonstrated by several classical databases. The results show that, in some situations, the proposed heuristic algorithm of attribute reduction can boost reduction quality of data to some extent.

关 键 词：粗糙集 属性约简 复合属性重要度 启发式算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]