变精度粗糙集的区域属性约简及其结构启发算法

摘要

采用二分类情形与三支决策区域, 研究变精度粗糙集(VPRS)两类属性约简及其结构启发算法.首先,依托三支决策区域构建分类区域, 提出分类区域保持(CRP)约简与决策区域保持(DRP)约简,得到对定性属性约简的量化扩张性, 设计基于核的结构启发算法;然后,研究两类区域约简的强弱关系, 设计由强至弱的结构启发算法, 得到二支决策拓展为三支决策的约简改进;最后,利用数据表实例与UCI数据集验证区域约简及其启发算法的有效性.%According to the two-category case and three-way decision regions,two types of attribute reductions for Variable Precision Rough Sets (VPRS) and their structural heuristic algorithms were studied.First of all,classification-regions were constructed by three-way decision regions,and Classification-Region Preservation (CRP) reduction and Decision-Region Preservation (DRP) reduction were proposed,quantitative expansion of the qualitative attribute reduction was obtained,and the structural heuristic algorithms based on cores were designed.Furthermore,the strong-weak relationships between the two kinds of regional reductions were studied,and structural heuristic algorithms from strong to weak were designed to achieve improvement from the two-way to three-way decisions.Finally,the validity of the relevant reductions and algorithms were verified by the data table and UCI data set.