一种基于可辨识向量的规则提取方法

Method of rule extraction based on discernible vector

针对信息表中的规则提取问题,应用粗糙集理论对其中的属性约简、属性值问题进行了研究,提出了一种基于可辨识向量的规则提取方法。根据粗糙集中的不可分辨关系建立了可辨识向量,利用可辨识向量的加法法则运算只需要对信息表扫描一次,就可以得到信息表的核属性集以及信息表的一个属性约简。在此基础上,利用条件属性与决策属性之间的对应关系,对信息表中的每条规则通过删除冗余属性值完成信息表的属性值约简,最终实现规则提取。数值实例和试验表明本算法是有效可行的。

To extract rules from information tables,the attribute reduction and attribute value reduction were investigated based on the theory of rough set in this study. A new rule extraction method was proposed based on discernible vector. According to the indiscernible relationship in a rough set,a discernible vector and its addition rule were defined. The core attribute set and the attribute reduction were then obtained by scanning the information table once using the discernible vector addition rules. Attribute value reduction was achieved through gradually deleting the redundant attribute values in every rule in the information table based on the correlation of condition attributes and decision attributes. Eventually,a concise rule set was obtained. The case study and experimental results demonstrate that the method is valid and sufficient for rule extraction.