基于遗传算法的决策形式背景的属性约简方法及其在决策分析中的应用  

Attribute Reduction Method for Formal Decision Contexts Based on Genetic Algorithm and its Application to Decision-making Analysis

作　　者：李金海[1] 梅长林[2] 张红英[2] 张晓[2] 

机构地区：[1]昆明理工大学理学院,昆明650500 [2]西安交通大学数学与统计学院,西安710049

出　　处：《小型微型计算机系统》2015年第8期1803-1808,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61305057)资助;昆明理工大学自然科学研究基金项目(14118760)资助

摘　　要：决策形式背景的最小约简可以使规则提取更加简便,也可以使所获取的规则更加紧凑,从而有利于数据的决策分析.对于如何快速求得决策形式背景的一个最小约简,已有一些启发式方法在这方面做了有益的尝试.然而,启发式思想求解最小约简遇到某些特殊的数据集会出现失效的现象.在决策形式背景中引入决策规则支持元与支持度,讨论了协调集与约简的等价判定定理,在此基础上提出基于遗传算法的决策形式背景的属性约简方法.数值实例分析表明,新约简方法能够在一定程度上弥补启发式算法存在失效现象的不足,从而有利于提高决策分析的效率.A minimal reduct of a formal decision context can make rule acquisition easier and the derived rules more compact, thereby beneficial to decision-making analysis of the data. The existing heuristic reduction methods have made a helpful attempt on how to effi- ciently find a minimal reduct of a formal decision context. However,these heuristic reduction methods will become invalid when being applied to special databases. In this paper,the notions of support elements and support degree of a decision rule are defined in formal decision contexts and are used to explore a theorem of judging whether a given conditional attribute set is a consistent set or even a re- duct. Then,a genetic-algorithm-based attribute reduction method is proposed for formal decision contexts and a numerical example is employed to demonstrate that the new reduction method can make up to some extent the invalidity of the existing heuristic reduction algorithms, thereby making more effective decision analysis of the data.

关 键 词：形式概念分析 决策形式背景 属性约简 遗传算法 决策分析 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]