基于扫描向量的属性约简方法

A method for attributes reduction based on scan vector

针对粗糙集理论中属性约简问题,提出了一种基于扫描向量的属性约简方法.根据粗糙集理论知识,定义了一个新概念——差别向量,利用差别向量将信息表转换成差别向量组;根据差别向量的结构特征,定义了差别向量加法法则;运用这个加法法则仅需对差别向量组扫描一次,就可以形成结构简洁却能代表原信息表属性特征的扫描向量.以扫描向量中的属性频率项作为属性约简搜索的启发信息,提高了属性约简效率.数值实例及数据库测试的结果表明该属性约简算法是有效可行的.

In order to deal with attributes reduction, one of the major problems in rough set theory, an attributes reduction algorithm was proposed based on scan vector, and a new conception of discernible vector was defined by which the information table can be transformed into discernible vector sets. Depending on the structural feature of the discernible vector, a plus rule for the discernible vector sets was defined, and a scan vector with concise structure but representing the information table can be obtained through scanning the discernible vector just one time. The item of attribute frequency in the scan vector was taken as heuristic information to improve the efficiency of attributes reduction. An illustration and experimental results indicate that the method proposed is much more effective.