Assuming a resonant origin of the quasiperiodic oscillations observed in the X-ray neutron star binary systems, we apply a genetic algorithm method for selection of neutron star models. It was suggested that pairs of kilo-Hertz peaks in the X-ray Fourier power density spectra of some neutron stars reflect a non-linear resonance between two modes of accretion disk oscillations. In several specific models, the two modes are related to physically plausible combinations of Keplerian, vertical and radial frequencies of geodesic orbital motion. We investigate this concept for a specific neutron star source, a fixed pair of modes and various neutron star equations of state. Each neutron star model is characterized by the equation of state (EOS), rotation frequency ($\Omega$) and central energy density ($\rho_\mathrm c$). These determine the spacetime structure governing geodesic motion and position dependent radial and vertical epicyclic oscillations related to the stable circular geodesics. When the parameters of neutron star model are fixed, the two considered modes imply a frequency-frequency relation which can be compared to the observation in order to eliminate the unsatisfactory sets (KR,$\rho_\mathrm c, \Omega$, EOS). For the elimination we use the advanced genetic algorithm. Genetic algorithm comes out from the method of natural selection when subjects with the best adaptation to assigned conditions have best chances to survive. The chosen genetic algorithm with sexual reproduction contains one chromosome with restricted lifetime, uniform crossing and genes of type 3/3/5. For encryption of physical description (KR,$\rho_\mathrm c, \Omega$, EOS) into chromosome we use the Gray code. As a fitness function we use correspondence between the observed and calculated pairs of eigenfrequencies.