The paper gives a detailed description of the principle and method of artificial neural network in the optimization of mining structure parameters. The mining technological conditions of the Gaofeng Mine are described. The selected input parameters include the thickness and dip angle of the orebody, the stability of rockmass and the layout of stopes. The selected output parameters include the dimensions of rooms and pillars and the height of sublevels. Therefore, a three-layer structured artificial neural network is created. 24 cases of mining methods are used as training samples and six practical cases are used for validation, which verify the feasibility and applicability of the method.