Aiming at the problem of large errors in the theoretical formula and experimental results of axial piston pump efficiency under different working conditions, a pump efficiency calculation method based on machine learning is proposed. Firstly, a nonlinear regression model based on axial piston pump efficiency is established, and the validity of the regression model is verified by the significance test; subsequently, a set of piston pump efficiency prediction model based on BP neural network is designed, and the prediction model is trained and verified based on the experimental data; finally, the prediction accuracies of the two models are evaluated respectively. The results show that, compared with the existing theoretical formula of piston pump efficiency under the conditions of variable pressure, speed and flow rate, the established nonlinear regression model and BP neural network model can significantly improve the prediction accuracy of the piston pump efficiency, and the average relative errors of the efficiency of the two methods are improved from 8.89% to 1.4% and 0.62%, respectively.