The accurate design and modification of cathode is one of the key problems in electrochemical machining （ECM）. In this paper, a digital cathode modification model is established based on the improved BP neural network, which substitutes a digital modification, changes the conventional manual correction method and remarkably increases the efficiency of the cathode modification process. In this study, the correction value of the cathode surface is predicted according to the data of multiple cathode modifications. The results show that the proposed model is effective in cathode modification, with a maximum absolute error between the predicted correction value and the actual one of only about 0. 015 mm, and that it can be widely used to the digital cathode modification of complex surfaces such as aeroengine blade, with greatly-reduced modification time and modification period as well as an improved electrochemical machining accuracy.