基于容错粗集理论的知识约简

粗集理论(RoughSetsTheory,RST)在处理不完全、不精确、不确定性智能信息问题等许多领域已取得成功的应用。然而RST理论还存在某些局限性,比如对等价关系的定义使得知识约简相对复杂化。该文提出利用容错粗集犤1犦RST的概念,建立属性值之间的容错关系,用容错关系对对象进行分类。该方法首先采用遗传算法求出属性阈值的最优组合,然后分别建立属性和对象的特殊矩阵,最后求出属性的相对约简和对象的相对商集。采用容错粗集进行知识约简减少了量化的过程,从而提高知识约简的有效性。该文最后提供了一个实例表明该方法是可行的。

Knowledge Reduction Based on Tolerance Rough Sets Theory

Rough Sets Theory has been developed in many application fields,which is successfully used to dispose such problems :as fragmentary data,incomplete,imprecise and uncertaint intelligent information.However there are some limits in the normal rough set theory.For instance,the equivalence relation is defined by the equal relation,which results in the complexity of the knowledge reduction.The paper proposes using the concept of tolerance rough sets theory to build objects' tolerance relation that can correctly classify objects.First we can use genetic algorithms to search for the optimal threshold set,then we build special matrix for attributes and objects.Finally we can get the relative reduction of attributes and relative absorbent set of objects.The method of using tolerance rough sets reduces the qualitative process,furthermore improves the validness of knowledge reduction.We present examples to illustrate our approach.