.png){kind=icon}
Return
Semi-supervised attribute reduction via attribute indiscernibility
.png){kind=icon}
PRE
AI
DOI：10.1007/s13042-022-01708-2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Attribute reduction based on rough sets plays an important role in data preprocessing. Discernibility pair, as an effective information measurement, has received extensive attention in attribute reduction. Unfortunately, the existing attribute importance measurement strategies based on discernibility pairs do not apply well to partially labeled data. Meanwhile, most of the existing attribute reduction algorithms focus on the relationships between objects and neglect the relationships between attributes, which may bring highly redundant attributes. Under the background of rough set theory, this paper studies the issue of semi-supervised attribute reduction, i.e. attribute reduction for partially labeled data. Firstly, we introduce the concept of discernibility pair based on object indiscernibility and propose a semi-supervised attribute reduction algorithm via the maximum discernibility pair by combining supervised and unsupervised discernibility pair strategies. Secondly, considering the relationships between attributes, we put forward new methods to define the similarity and distinction between attributes by discernibility pairs. Thirdly, we propose a semi-supervised attribute reduction algorithm by indiscernible attribute classes. Finally, comparative experiments indicate that the proposed algorithms are effective.
Keywords:
Rough sets
Semi-supervised attribute reduction
Discernibility pair
Indiscernible attribute class
Journal
IF:
2.7
Papers:
3.2K
Citations:
5.6K
Organization
Cited Papers
.png){kind=icon}
ShareSave
ShareSaveShareSave
View more