Assuming a resonant origin of the twin peak quasiperiodic oscillations observed in the X-ray neutron star binary systems, we apply a genetic algorithmmethod for selection of neutron star models. Each neutron star model is characterized by the equation ofstate (EOS), rotation frequency and central energy density. These determine the spacetime structure governing geodesic motion and position dependent radial and vertical epicyclic oscillations related to the stable circular geodesics. For the eliminationwe use the advanced genetic algorithm. As a fitness function we use correspondence between the observed and calculated pairs of eigenfrequencies.