动态粗糙模糊集及其在模糊规则提取中的应用  

Dynamic Rough Fuzzy Sets and its Application in Extracting Fuzzy Rules

作　　者：程昳[1,2,3,4] 苗夺谦[1,3,4] 冯琴荣[1,3,4] 

机构地区：[1]同济大学嵌入式系统与服务计算教育部重点实验室,上海201804 [2]中国海洋大学数学科学学院,山东青岛266071 [3]同济大学计算机科学与技术系,上海201804 [4]国家高性能计算机工程中心同济分中心,上海201804

出　　处：《小型微型计算机系统》2009年第2期289-294,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60475019;60775036)资助;2006博士学科点专项科(20060247039)资助

摘　　要：基于粗糙模糊集的规则提取方法通常分为两步:首先利用粗糙模糊集进行属性约简,然后采用提取模糊规则的方法提取规则.在规则提取的预处理阶段通过属性约简某种程度上可以缩短规则提取的时间,但其固有的不足导致不利于产生良好的规则.在模糊规则产生过程中避开属性约简,可以提高规则提取方法的适用性,降低计算复杂度.本文提出了动态粗糙模糊集的概念,基于此的规则提取算法不再依赖于属性约简,而是基于粒度序和逐步缩小的论域.首先,通过两种不同方式定义了动态粗糙模糊集并得到一些重要性质;在此基础上提出一种新的模糊规则提取算法;最后通过对比实验说明了算法的有效性.Most rule induction algorithms based on rough fuzzy sets theory often include two steps： attributes reduction based on rough fuzzy sets and fuzzy rules induction based on conventional rules mining algorithms. It＇s s useful to shorten time to some extent by attributes reduction in preprocessor of rule mining. However, attributes reduction may make against the induction of fine rules due to the flaws of itself. Avoiding the process of attributes reduction in fuzzy rules induction permits to improve the adaptability of generating fuzzy rules and reduce computational complexity. In this paper, the dynamic rough fuzzy sets were presented. A rule induction algorithm which, different from most known fuzzy rules induction , is not based on attributes reduction but granulation order and dwindle universe was designed. Firstly, the dynamic rough fuzzy sets were defined in two different ways and some important properties were obtained, creating a base for induction of fuzzy rules. Secondly, an algorithm, based on dynamic rough fuzzy sets was put forward for decision rule mining. At last, application of the algorithm was illustrated by an example. The results showed that the algorithm was effeetive, supported by comparisons to the application of fuzzy rules induction based on attributes reduction.

关 键 词：动态粗糙模糊集 决策表 模糊规则 粒度序 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]