Comparative analysis of differential evolution methods to optimize parameters of fuzzy classifiers
- Authors: Mekh M.A.1, Hodashinsky I.A.1
-
Affiliations:
- Tomsk State University of Control Systems and Radioelectronics
- Issue: Vol 56, No 4 (2017)
- Pages: 616-626
- Section: Control in Stochastic Systems and Under Uncertainty Conditions
- URL: https://journals.rcsi.science/1064-2307/article/view/219937
- DOI: https://doi.org/10.1134/S1064230717040116
- ID: 219937
Cite item
Abstract
We have compared the efficiency of 14 modifications of the differential evolution method for optimizing the parameters of fuzzy classifiers, the rule bases of which are initialized by the algorithm of the structure generation on the basis of extreme parameter values. The comparison was conducted on 12 well-known datasets of the KEEL repository. Based on the results of the operation of classifiers optimized by these algorithms, we ranked the methods and compared their analogs. The comparison criteria are represented by the percentage of correct classification and the number of rules in a classifier.
About the authors
M. A. Mekh
Tomsk State University of Control Systems and Radioelectronics
Email: hodashn@rambler.ru
Russian Federation, Tomsk, 634050
I. A. Hodashinsky
Tomsk State University of Control Systems and Radioelectronics
Author for correspondence.
Email: hodashn@rambler.ru
Russian Federation, Tomsk, 634050