基于属性重要性的决策树规则提取算法

对属性进行约简为从关系数据库中挖掘简洁的规则创造了条件。但从属性约简后的数据集中提取最简规则的问题仍是一个NP难题,一般借助于启发式算法。提出了一种新的基于属性重要性的规则提取算法,称为IADT(importantattributedecisivetree)算法,采用粗糙集理论中的属性重要性概念,通过建立树结构来提取规则。算法可以避免面对NP难题,获得相对简单的规则。计算实例表明,IADT算法具有良好的实用性,有进甚至可以获得最简规则。通过与ID3算法进行的比较表明,IADT算法为规则树算法提供了一种新的属性选择标准。

Rule abstracting algorithm by decisive tree based on the importance of attribute

Though it is possible to get more brief rules with rdeuced attributerss, getting the briefest rules from the reduced data set is a NP hard question and depends on the heuristic algorithm. In this paper, with the help of the idea of the importance of attribute in rough set theory, a new building of a rule tree algorithm named IADT(Important Attribute Decisive Tree) is developed. The algorithm can avoid the NP hard question and get relative brief, maybe the briefest, rules and can be easily used in compare with ID3, IADT gives a new method of selecting attributes in building a rule tree.