多半径邻域粗糙集改进约简算法

属性约简是粗糙集理论中的重要问题。许多学者针对邻域粗糙集提出多种属性约简方法，包括应用最为广泛的启发式算法。在多半径邻域粗糙集的基础上，针对当前启发式约简算法往往会包含一定冗余属性的缺陷，提出一种融合属性权重影响的改进约简运算方法，通过根据各属性权值大小设置阈值使得约简结果能够消除冗余属性。实验选取UCI的数据集与当前几种常用启发式约简算法进行比较分析。实验结果表明，所提出的属性约简方法能够得到更优的约简集合，同时更大程度地保留了决策表本身的知识信息，具有较高的分类能力。

Improvement to attribute reduction algorithm in neighborhood rough set

Attribute reduction is one of important aspects in rough set theory. Scholars have proposed series of attribute reduction methods in neighborhood rough set, including the widely applied heuristic algorithm. To deal with the shortage of existing reduction algorithm containing redundant attributes, an attribute reduction method based on weights approach is proposed on the basis of neighborhood rough set with multiple radius. Thresholds are set up to eliminate redundant attributes in reduction results according to weights of every condition attribute. Several databases are applied to analyze algorithm performance. The experimental results show that this method can retain more knowledge and information of decision table and has better performance.