Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
495154 | Applied Soft Computing | 2015 | 13 Pages |
Abstract
- A hybrid method is proposed for efficient subset selection in high-dimensional datasets. The symmetrical uncertainty (SU) criterion is exploited to weight features in filter phase.
- In wrapper phase, both fuzzy imperialist competitive algorithm (FICA) and Incremental Wrapper Subset Selection with replacement (IWSSr) in weighted feature space are executed to search and find relevant attributes.
- The proposed method has been assessed by applying on 10 standard high-dimensional datasets. We compared our proposed algorithm with other five hybrid algorithms (LFS, IWSS, IWSSr, BARS, and Grasp) and two filter methods (FCBF and PCA). The comparison between the results of our method and others confirms that our method has the best accuracy.
- The average number of attributes selected by proposed algorithm is considerably less than the other methods.
- The diagrams show low convergence time and low number of iterations with regard to other methods.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science Applications
Authors
Mostafa Moradkhani, Ali Amiri, Mohsen Javaherian, Hossein Safari,