Abstract: When solving optimization problem, the traditional method determines the objective function on the basis of the least squares principle without considering the influence of the error of the original measurement data on the results and the accuracy of calculation. Therefore machine learning algorithm is proposed in this paper to improve the general objective function and to solve the hydrogeological parameters in combination with particle swarm optimization of double evaluation. The result shows that machine learning algorithm has good convergence and stability as well as high solution efficiency, is simple and easy to realize.