摘要
在粗糙集模型中,α量化不可分辨关系是强与弱不可分辨关系的推广形式.然而值得注意的是,基于这三种不可分辨关系的粗糙集并未考虑数据中属性的测试代价.为解决这一问题,提出了测试代价敏感的α量化粗糙集模型,从二元关系的角度使得粗糙集模型代价敏感,并将新模型与基于强不可分辨、弱不可分辨以及传统α量化不可分辨关系的粗糙集模型进行了对比分析.进一步地,通过分析传统启发式算法在求解约简的过程中未考虑降低代价这一不足之处,提出一种新的属性适应性函数,并将其应用于基于遗传算法的约简求解中.实验结果表明该方法不仅可以降低由边界域所带来的不确定性而且同时降低了约简后的测试代价.
in rough set model, α quantitative indiscernibility relation is a generalization ot both strong and weak indiscernibility relations. However, such three indiscernibility relations based rough sets do not take the test costs of the attributes into consideration. To solve this problem, α test-cost-sensitive quantitative indiscernibility relation based rough set is proposed. From the viewpoint of the binary relation, the new rough set is then sensitive to test costs. Moreover, the relationships among strong, weak, α quantitative and test-cost-sensitive α quantitative indiscernibility relations based rough sets are explored. Finally, it is noticed that the traditional heuristic algorithm does not take the decreasing of cost into account. Therefore, not only a new fitness function is proposed, but also such fitness futtction is carried out in genetic algorithm for obtaining reduct with minor test cost. The experimental results show that such approach not only decreases the uncertainty comes from boundary region, but also decreases the cost of reduct.
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2017年第1期228-240,共13页
Systems Engineering-Theory & Practice
基金
国家自然科学基金(61572242
71671086
61473157)
江苏省普通高校研究生科研创新计划项目(KYLX16_0021)~~
关键词
粗糙集
α量化不可分辨关系
测试代价敏感
属性约简
rough set
α quantitative indiscernibility relation
test-cost-sensitive
attribute reduction
