相对知识粒度序决策表的属性约简方法

摘要

Knowledge reduction has been paid more and more attention in ordered decision table. It proposes a new relative knowledge granulation in ordered decision table based on dominance classes and knowledge granulation to measure relative uncertainty of an attribute set in ordered decision table, then defines an relative atribute significance and designes a heuristic attribute reduction algorithm of ordered decision table. The validity of this algorithm is analyzed and verified by an example.%序决策表中的知识约简越来越受到关注,在优势类和知识粒度的基础上,引入了序决策表中的一种知识相对粒度,度量了属性集在序决策表中的相对不确定性,进而给出了属性相对重要度的定义.并设计了序决策表的一种启发式属性约简算法,通过实例分析和验证了算法的有效性.