In order to improve prediction accuracy of blasting vibration velocity,a model based on relevance vector machine(RVM) was proposed.The nonlinear mapping relationship between blasting vibration velocity and its influencing factors was established using the model.Three main factors affecting blasting vibration velocity(explosive charge,distance and elevation difference) were fitted for training by 36 sets of generated data based on which the remaining 5 samples were accurately predicted.The model was applied as an example and compared with the prediction results of BP neural network model and GA-BP neural network model.Under the same influence factor and data sample conditions,RVM model has higher prediction accuracy and lower dispersion.Compared to the actual values,the average relative errors of blasting vibration velocity predicted by RVM are obviously lower than those predicted by BP neural network and GA-BP neural network,which further verifies that the RVM model can improve the accuracy of the prediction accuracy and stability.