The influence of blasting vibration to surroundings around the blasting area can not be ignored, in order to guarantee the safety of surroundings around blasting area, blasting vibration forecasting model based on neural network is established by improved BP neural network in this paper. The inputs of the model are the largest single fire dynamite, height difference and horizontal distance between blasting source and testing point or protect object, and the outputs are vibration velocity and vibration frequency in three directions. At last, the model is applied to a mine, and the results show that it is very reliable for the method to forecast the blasting vibration, which has a certain practical significance for guiding blasting design and construction.