원문정보
초록
영어
Numerous studies have focused on feature selection using many algorithms, but most of these algorithms encounter problems when the amount of data is large. In this paper, we propose an algorithm that handles a large amount of data by partitioning the data to process a reduction, and then selecting the intersection of all reducts as a stable reduct. This algorithm is successful but may suffer from loss of information if the samples are unsuitable. The proposed algorithm is based on discernibility matrix and function. Furthermore, the method can address the case in which the data consist of a significant amount of information. Our results show that the proposed algorithm is powerful and flexible enough to successfully target a range of different domains and can effectively reduce computational complexity as well as increase reduction efficiency. The efficiency of Proposed Algorithm is illustrated by experiments with UCI datasets further.
목차
1. Introduction
2. Related Work
3. Rough Set Base Approach
3.1. Information System and Indiscernibility Relation
3.2. Class Approximation
3.3. Dispensable and Indispensable Features
3.4. Reduct and CORE
3.4. The Discernibility Matrix and the Discernibility Function
4. Minimal Attribute Reduction Based on Discernibility Function
5. Feature Selection Using Rough Set
6. Selection Minimal Attributes Reduction
6.1. Proposed Algorithm
6.2. How Does The Algorithm Work?
6.3. Illustrative Example
7. Algorithm Testing and Comparison (Implementation)
8. Conclusion
References