基于投票式属性重要度的快速属性约简算法

Fast Attribute Reduction Algorithm Based on Importance of Voting Attribute

作为经典Pawlak粗糙集的扩展,邻域粗糙集能有效处理数值型的数据。但是,因为引入了邻域粒化的概念,所以邻域实数空间下的计算量要比经典离散空间下的计算量大得多。对于邻域粗糙集算法而言,能够有效且快速地找到数据集的属性约简是十分有意义的。为此,针对现有算法中属性重要度定义的不足,首先提出了一种改进的投票式属性重要度,然后进一步提出了一种基于投票式属性重要度的快速属性约简算法。实验证明,与现有算法相比,在保证分类精度的前提下,该算法能更快速地得到属性约简。

As an extension of the classical Pawlak rough set,neighborhood rough sets can efficiently manipulate numerical data.However,because the concept of neighborhood granulation is introduced,computational complexity in the neighborhood real space is much larger than that in the classical discrete space.For the neighborhood rough set algorithm,it is very meaningful to find the attribute reduction of the data set efficiently and quickly.To this end,an improved definition of voting attribute importance was proposed for the shortcomings of the definition of attribute importance in existing algorithms,then a fast attribute reduction algorithm based on importance of voting attribute was proposed.Compared with the existing algorithms,the experiment proves that the algorithm can get the attribute reduction more quickly under the premise of ensuring the classification accuracy.