1. Accelerated multi-granularity reduction based on neighborhood rough sets.
- Author
-
Li, Yizhu, Cai, Mingjie, Zhou, Jie, and Li, Qingguo
- Subjects
ROUGH sets ,GRANULAR computing ,COMPUTATIONAL complexity ,MATHEMATICAL models - Abstract
The notion of multi-granularity has been introduced into various mathematical models in granular computing. For example, neighborhood rough sets can derive a good multi-granularity structure by gradually changing the size of neighborhood radius. Attribute reduction is an important topic in neighborhood rough sets. However, in the case of multi-granularity, the challenge of high computational complexity and difficulty in synthesizing multi-granularity information when performing reduction algorithms always exists. To address such limitations, an accelerated algorithm for multi-granularity reduction is designed. Firstly, we construct a multi-granularity reduction structure with multiple different neighborhood radii to reduce the elapsed time of computing reducts. In this way, the consumed time of calculating the distance is similar to the one of single granularity reduction, and the elapsed time of computing multi-granularity reducts can be reduced. Secondly, multiple granularity information is integrated in each attribute evaluation. Finally, we evaluated the proposed method from multiple perspectives on 12 UCI datasets. Compared with other multi-granularity reduction algorithms, the proposed method not only generates reducts with relatively high quality, but also improves the time efficiency of multi-granularity reduction algorithm. [ABSTRACT FROM AUTHOR]
- Published
- 2022
- Full Text
- View/download PDF