Ilya A. Hodashinsky, Ivan V. Gorbunov, Konstantin S. Sarin, Sofiya R. Subkhankulova

Tomsk State University of Control Systems and Radioelectronics

The algorithm of feature selection was proposed based on mine blast optimization. Two algorithms are described for a structure generation of fuzzy approximators. The first of them is dynamic partitioning of the input space and the second is piecewise linear initialization. An algorithm are described for structure generation of a fuzzy classifier based on the extreme values of observation table. The results of the algorithms benchmarked on real data from the repository KEEL

Takagi-Sugeno fuzzy systems, structural identification, mine blast algorithm, piecewise linear initialization, dynamic partitioning of the input space, fuzzy classifier