Loading…
Cost-sensitive three-way class-specific attribute reduction
The theory of rough sets provides a method to construct three types of classification rules, leading to three-way decisions. From such a point of view, we introduce the concept of cost-sensitive three-way class-specific attribute reducts. Based on the semantics of the three-way decisions, we introdu...
Saved in:
Published in: | International journal of approximate reasoning 2019-02, Vol.105, p.153-174 |
---|---|
Main Authors: | , |
Format: | Article |
Language: | English |
Subjects: | |
Citations: | Items that this one cites Items that cite this one |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | The theory of rough sets provides a method to construct three types of classification rules, leading to three-way decisions. From such a point of view, we introduce the concept of cost-sensitive three-way class-specific attribute reducts. Based on the semantics of the three-way decisions, we introduce a monotonic result cost in decision-theoretic rough set model, called the result cost of three-way decisions. We provide a critical analysis of classification-based attribute reducts from result-cost-sensitive and test-cost-sensitive perspectives. On this basis, we propose class-specific cost-sensitive attribute reduction approaches. More specifically, we define a class-specific minimum cost reduct. The objective of attribute reduction is to minimize result cost and test cost with respect to a particular decision class. We design two algorithms for constructing a family of class-specific minimum cost reducts based on addition–deletion strategy and deletion strategy, respectively. The experimental results indicate that the result cost of three-way decisions is monotonic with respect to the set inclusion of attributes and the class-specific minimum cost reducts can make a better trade-off between misclassification cost and test cost with respect to a particular decision class. |
---|---|
ISSN: | 0888-613X 1873-4731 |
DOI: | 10.1016/j.ijar.2018.11.014 |