Ilya A. Hodashinsky (hodashn@rambler.ru), Alexander E. Anfilofiev (yowwi00@gmail.com), Marina B. Bardamova (e-mail: 722bmb@gmail.com), Sergey S. Samsonov (723_sss@fb.tusur.ru), Igor V. Filimonenko (ifilimon96@mail.ru)

Tomsk State University of Control Systems and Radioelectronics

The feature selection is an NP-hard problem, it is guaranteed the optimal solution can be found only by a full search. In the article, we describe the approach to feature selection based on ranking and cross-validation. For the formation of optimal feature sets, binary meta-heuristic algorithms are used: gravitational search algorithm, weed algorithm, monkey algorithm and krill herd algorithm.

cross-validation, feature selection, classifiers, binary metaheuristics