• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2015, Vol. 37 ›› Issue (11): 2169-2174.

• 论文 • 上一篇    下一篇

因素空间的属性圆定义及其在对象分类中的应用

崔铁军1,2,3,马云东3   

  1. (1.辽宁工程技术大学安全科学与工程学院,辽宁 阜新 123000;2.矿山热动力灾害与防治教育部重点实验室,辽宁 阜新 123000;
    3.大连交通大学辽宁省隧道与地下结构工程技术研究中心,辽宁 大连 116028)
  • 收稿日期:2014-10-20 修回日期:2014-12-19 出版日期:2015-11-25 发布日期:2015-11-25
  • 基金资助:

    国家自然科学基金资助项目(61350003);教育部高校博士学科点专项科研基金资助项目(20102121110002)

Definition of attribute circle in factor space and its
application in object classification 

CUI Teijun1,2,3,MA Yundong3   

  1. (1.College of Safety Science and Engineering,Liaoning Technical University,Fuxin 123000;
    2.Key Laboratory of Mine Thermodynamic Disasters and Control of Ministry of Education,Fuxin 123000;
    3.Tunnel & Underground Structure Engineering Center of Liaoning,Dalian Jiaotong University,Dalian 116028,China)
  • Received:2014-10-20 Revised:2014-12-19 Online:2015-11-25 Published:2015-11-25

摘要:

为使汪培庄先生提出的因素空间理论便于应用,和基于该理论对多域值属性影响对象集合进行聚类分析,提出了以研究对象为中心的图形化域值属性表示方法,即属性圆。属性圆可以表示无穷多个域属性对对象的影响。先基于属性圆概念进行对象的相似性分析,后为计算方便将图形定义转化为数值相似性定义,进而研究了对象集合的聚类分析方法。实施的聚类原则为:严格遵照相似与不相似划分,参考模糊相似划分。列举了一个实际电气系统的系统可靠性表述群作为研究对象集合,对表述群进行聚类分析。结果表明:决策集D与对象集U的对应关系说明对对象集的划分就其决策属性而言是非奇异的、准确的。这说明尽管在不同环境下对系统进行了可靠性评价,但是这些评价语义是相对客观的,评价的语义可以相互佐证。

关键词: 因素空间, 多域值属性, 属性圆, 对象分类

Abstract:

In order to achieve easier application of the factor space theory easily based on its impact on the object set of multidomain value attributes,we propose a graphical representation for an object with the domain value attributes, namely attribute circle, using cluster analysis.The circle can represent the impact of infinite multidomain attributes on the object.We first analyze the similarities of the objects based on the attribute circle concept.For the convenience of calculation,graphical definition is converted to numerical similarity definition, and the clustering analysis method of object set is studied.The clustering principles are: strictly complying with similar division or dissimilar division,and referencing to the fuzzy similarity division.We take the system reliability of an electrical system as the research object set,and analyze the cluster of the expression group.The corresponding relationship between the decision set D and the object set U show that the division of the object sets is singular and accurate in terms of decision attributes.Although the system reliability is evaluated in different circumstances,these evaluation semantics are relatively objective and they can support each other.

Key words: factors space;multi-domain value attribute;attribute circle;object classification