In order to improve the dynamic and static characteristics of computer numerical control (CNC) milling machine more effectively, the combination of experimental design and mathematical statistics method was adopted for the structure size optimization of XK719 CNC milling machine. The CNC milling machine using Latin hypercube experimental design method to the sampling, on the basis of using the moving leastsquare method to fitting the sample point, the approximation function was constructed. The approximate function was optimized by using the adaptive response surface method. By 25 iterations, under the condition that changes in mass of CNC milling machine was less than 3%, the first order natural frequency was implemented to increase 163%, and the second order frequency to increase 13%. The results show that the method can implement for the rapid optimal design of machine tool.