Binarization of the Swallow Swarm Optimization for Feature Selection
| Autor: | Artem Slezkin, Alexander Alexandrovich Shelupanov, Ilya Hodashinsky |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Fuzzy rule
business.industry Computer science Swarm behaviour Pattern recognition Feature selection ComputingMethodologies_PATTERNRECOGNITION Feature (computer vision) Algebraic operation Classifier (linguistics) Relevance (information retrieval) Artificial intelligence business Software Merge (linguistics) |
| Zdroj: | Programming and Computer Software. 47:374-388 |
| ISSN: | 1608-3261 0361-7688 |
| DOI: | 10.1134/s0361768821050066 |
| Popis: | In this paper, we propose six methods for binarization of the swallow swarm optimization (SSO) algorithm to solve the feature selection problem. The relevance of the selected feature subsets is estimated by two classifiers: a fuzzy rule-based classifier and a classifier based on k-nearest neighbors. To find an optimal subset of features, we take into account the number of features and classification accuracy. The developed algorithms are tested on datasets from the KEEL repository. For the statistical evaluation of the binarization methods, we use Friedman’s two-way analysis of variance by ranks for related samples. The best feature selection result is shown by a hybrid method based on modified algebraic operations and MERGE operation introduced by the authors of this paper. The best classification accuracy is achieved with a V-shaped transfer function. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink