In view of the difficulty in the selection of technological parameter in power-mixed pulse electrochemical polishing, a scheme based on neural network for the selection of technological parameter was put forward. The model of Surface toughness Forecasting was established. After being compared with experiment data, it was found that the experiment results and expected results share much similarity. It indicates that the system is able to predict the roughness of surface accurately under a given condition. Therefore, it can cast some light to practical power-mixed pulse lectrochemical polishing.