基于矩阵的可变粒度变精度邻域粗糙集近似集更新方法

Matrix-based Approaches for Updating Approximations in Variable Granulation Variable Precision Neighborhood Multigranulation Rough Sets While Neighborhood Classes Decreasing

邻域粗糙集可以同时处理名义与数值属性,多粒度粗糙集提供多个粒度视角下的目标概念近似,变精度粗糙集使得近似集计算不再局限于完全包含。本文首先提出了一种同时具有以上三种粗糙集模型长处并且粒度可变的变精度多粒度邻域粗糙集模型,并设计基于矩阵的近似集计算与更新方法:首先提出静态计算近似集的矩阵算法,继而考虑在邻域粒变小时,基于静态计算算法对近似集进行更新,提出一种邻域粒变小时近似集更新的矩阵算法,最后通过UCI公开数据集实验验证了计算与更新算法的有效性。

Neighborhood rough sets can handle category and numerical attributes simultaneously. Multigranulation rough sets provide approximations of target concepts from multiple granulation perspectives. Variable precision rough sets make the calculation of approximate sets no longer limited to complete inclusion. This paper first proposes a Variable Granulation Variable Precision Neighborhood Multigranulation Rough Sets(VG-VP-NMGRS) with the advantages of the above three rough set models and variable granulation. And propose VG-VP-NMGRS approximation set calculating and updating method of matrix: firstly, a relative matrix-based algorithm for calculating approximations in VG-VP-NMGRS is proposed, and then when the neighborhood classes are decreasing, the incremental algorithm for updating approximations is proposed based on the static calculating algorithm. Finally, the validity of the static and incremental algorithms is verified by UCI data set experiments.