摘要
多粒度是近年来粗糙集领域研究的一个热点方向,而粒度约简是其中的一个核心问题。为了使得多粒度粗糙集能够用于处理连续型数据,引入模糊概念,构建了基于模糊等价关系的悲观多粒度模糊粗糙集模型,并进一步给出了粒度重要度的度量方法,设计一种基于启发式的粒度约简算法。以UCI(University of California Irvine)中3组数据集进行分析,实验结果表明所设计的算法能够在保持分类准确率不发生较大变化的情况下约去冗余的粒结构。
Multigranulation is one of the hot directions in rough set theory, while granulation reduction is a key prob lem in multigranulation rough set. To deal with the continuous data with multigranulation rough set, the fuzzy concept is employed and the pessimistic multigranulation fuzzy rough set is constructed based on fuzzy equivalence relation. Moreo- ver, the measurement of granulation significance is presented and a granulation reduction algorithm is designed with heu- ristic idea. Finally, the algorithm is tested on three UCI（University of California Irvine）data sets, the experimental re- suits show that the proposed algorithm can reduce redundant granulation struc hout the great changing of accura- cy of classification.
出处
《武汉理工大学学报》
CAS
CSCD
北大核心
2014年第8期133-137,共5页
Journal of Wuhan University of Technology
基金
国家自然科学基金(61100116)
江苏省自然科学基金(BK2011492)
徐州工程学院青年课题(xky2011201)
关键词
悲观多粒度
模糊相似矩阵
模糊决策信息系统
模糊粗糙集
粒度约简
pessimistic multigranulation
fuzzy similarity matrix
fuzzy decision information system
fuzzyrough set
granulation reduction
