High-speed milling has been widely applied in the manufacturing industry. The orthogonal experiment of multivariate factors L9(34) was designed in order to acquire processing characteristics of surface quality efficiently and entirely. Based on this, a grey neural network combinational model was established by combining small sample of grey theory, simple and super strong characteristics of neural network non-linearity. Experimental results show that the model has accurate fitting precision and perfect generalization ability, which will provide a way to further simulation research of high speed milling technics.