Modeling and control

Effectiveness investigation of adaptive evolutionary algorithms for data mining information technology design

M. E. Semenkina

M. F. Reshetnev Siberian State Aerospace University

Abstract: New uniform crossover operators that introduce the selective pressure on the stage of the recombination are suggested both for the genetic and genetic programming algorithms. Both algorithms’ performance improvement is demonstrated with benchmark problems. New method of evolutionary algorithms selfconfiguring based on the tuning the probability of genetic operators use is developed and implemented. Additionally, the neural networks automated design method based on genetic programming algorithm is suggested. Comparison with known analogues is fulfilled that demonstrates the high level performance of implemented algorithms.

Keywords: genetic algorithm, genetic programming, uniform crossover, self-configuration, symbolic regression, neural networks, classification.

Bibliographic databases:

•