一种改进的变论域属性约简算法

An Approach to Attribute Reduction Based on Developed Variable Universal

本文主要对Rough Set理论中的属性约简问题进行了研究,从变论域的角度出发,利用可辨识矩阵对不可分 辩类进行了定义,结合属性重要性及属性间依赖性的讨论,构造了一种变论域的启发式算法,并通过实例和UCI数据库证 明了该算法的有效性。

This paper mainly discusses the attribute reduction in the rough set theory. It begins with the variable universal theory,then uses the discernibility matrix to define the undistributed class, which constitutes a heuristic algorithm for attribute reduction. Experiments show that it can get better effect and is effectiue in the reduction.