To solve the modelling problem of wheel-rail force, a modelling approach based on the genetic algorithm and radial basis function neural network method is proposed, which can predict the output of wheel-rail force using the input of track irregularities. In order to improve the accuracy of the designed neural network and relieve the computational burden, the centers, widths and weights of the neural network are determined using the maximum distance measure, the least square method and genetic algorithm, respectively. The simulation results indicate that the proposed method can predict wheel-rail force with high precision.