智能系统学报2024，Vol.19Issue(6)：1468-1478,11.DOI:10.11992/tis.202309032
特定类的代价敏感近似属性约简
Cost sensitive approximate attribute reduction for specific classes
摘要
Abstract
Class-specific attribute reduction refers to reducing attributes that are provided specifically for a given de-cision class.Existing class-specific attribute reduction methods are often too strict,which limits their applicability in certain scenarios.For noisy data,this paper proposes a cost-sensitive approximate attribute reduction method tailored for specific classes.First,the method combines information from the positive and boundary regions to define the relative uncertainty for a specific class.Then,attribute importance is calculated using relative uncertainty and test cost,allowing for attribute selection based on importance and avoiding the inclusion of redundant attributes by relaxing the relative un-certainty.Finally,the study introduces a cost-sensitive approximate heuristic attribute reduction for specific classes.Ex-perimental results show that the proposed method can maintain or even improve the reduction quality while achieving a more streamlined reduction compared to other methods,with a relatively lower test cost for the reduction set.关键词
粗糙集/不确定信息/特定类/相对不确定度/属性重要度/测试代价敏感/近似属性约简/启发式算法Key words
rough set/uncertain Information/specific class/relative uncertainty/attribute importance/test-cost-sensit-ive/approximate attribute reduction/heuristic algorithm分类
信息技术与安全科学引用本文复制引用
胡军,黄小涵..特定类的代价敏感近似属性约简[J].智能系统学报,2024,19(6):1468-1478,11.基金项目
国家自然科学基金项目(62221005,62276038) （62221005,62276038）
重庆市自然科学基金项目(cstc2021ycjh-bgzxm0013) （cstc2021ycjh-bgzxm0013）
重庆市教委重点合作项目(HZ2021008). （HZ2021008）
