基于区分对的混合型弱标记数据增量约简算法  

Incremental Reduction Algorithm for Hybrid Weakly Labeled Data Based on Discriminating Pairs

作　　者：金莎 郑颖春[1] JIN Sha;ZHENG Yingchun(School of Science,Xi'an University of Science and Technology,Xi'an 710600,China)

机构地区：[1]西安科技大学理学院,陕西西安710600

出　　处：《河南科技大学学报（自然科学版）》2022年第3期92-99,M0008,共9页Journal of Henan University of Science And Technology:Natural Science

基　　金：国家自然科学基金项目(71473194);陕西省科技计划项目(2020CGXNG-013)。

摘　　要：针对混合型的弱标记不完备数据,提出属性变化的增量式属性约简算法。在混合型弱标记决策系统中,基于正域概念引入改进的属性区分关系定义,给出系统中属性动态变化时区分关系的增量式更新方法,提出了混合型弱标记决策系统下的启发式增量属性约简算法。实验结果表明:在处理大规模数据时,本文提出的增量属性约简算法相较于非增量约简算法能节约81.5%的时间,相比于仅利用无标记数据得到的约简结果,分类精度平均值提高了14.42左右。Most of the current attribute reduction algorithms only consider discrete labeled data.As a result,the classification performance is not high.In this paper,for the mixed weakly labeled incomplete data,an incremental attribute reduction algorithm with attribute changes was proposed.Based on the concept of positive domain,an improved attribute discrimination relationship definition was introduced in the hybrid weakly labeled decision system.The incremental update method of distinguishing the relationship was proposed when the attributes of the subsystem changed dynamically.In the hybrid weakly labeled decision system,a heuristic incremental attribute reduction algorithm was established.The result show that compared with the non-incremental reduction algorithm,the incremental attribute reduction algorithm proposed in this paper can save 81.5%of time.Compared with the reduction result obtained by only using unlabeled data,the average classification accuracy of the classification accuracy is increased by about 14.42.

关 键 词：粗糙集 属性约简 区分对 属性相对区分度 增量学习 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]