一种基于知识粒度的启发式属性约简算法

摘要

Attribute reduction is very important to knowledge acquisition in rough set theory. According to the consistency of attribute similarity and knowledge granularity, a new algorithm for attribute reduction based on knowledge granularity is proposed by calculating the similarity between condition attributes and decision attributes, as well as the similarity between condition attributes. The condition attributes are ordered descendingly based on the similarity between condition attributes and decision attributes, and the reduction set is obtained by selecting the important attributes based on the similarity between condition attributes. Theory analysis and the experiment results show that this algorithm reduces the calculation complexity and improves reduction effect.%属性约简是粗糙集理论进行知识获取的核心问题之一.根据属性相似度与知识粒度的一致性,通过条件属性与决策属性以及条件属性之间的相似度度量,提出了一种基于知识粒度的启发式属性约简算法.根据条件属性与决策属性的相似度对条件属性进行降序排列,根据条件属性之间的相似度度量选择重要的属性,从而得到约简集合.理论分析与实验结果表明,该算法具有较高的运行效率和较好的约简效果.