With the speed upgrade of the high-speed train, the aerodynamic drag becomes one of the key factors to restrain the train speed and energy saving. In order to reduce the aerodynamic drag of train head, a new parametric approach called local shape function (LSF) was adopted based on the free form surface deformation (FFD) method and a new efficient optimization method based on the response surface method (RSM) of GA-GRNN. The optimization results show that the parametric method can control the large deformation with a few design parameters, and can ensure the deformation zones smoothness and smooth transition of different deformation regions. With the same sample points for training, GA-GRNN performs better than GRNN to get the global optimal solution. As an example, the aerodynamic drag for a simplified shape with head + one carriage + tail train is reduced by 8.7%. The proposed optimization method is efficient for the engineering design of high-speed train.