关键词: 概念层次, 描述子, 属性约简, 优化可信规则, 粗糙集

Abstract: The concept hierarchies provide a new way to solve such problems as the data are increasing too fast, there are so many rules, and different decision makers need obtain some rules from different hierarchies. In the context that every condition attribute had a concept hierarchy, the relations of positive regions and rules from different levels were analyzed based on rough set theory. Then, the algorithm of optimal credible rules based on rough sets in concept hierarchies was proposed, which was along with concept hierarchies from top to down. The algorithm improved the existing reduction strategy, realized the reduction and obtained the optimal credible rules by descriptors. Besides, it also considered there was null or not a new object in the positive region, and the effectiveness of extracting rules was improved. Finally, the feasibility of this algorithm was validated by an example.

Key words: concept hierarchies, descriptor, attribute reduction, optimal credible rule, rough set