Abstract:Aiming at the limitations of wavelet neural network, we use particle swarm algorithm to optimize the wavelet neural network. On this basis, a fitting model of GPS elevation abnormality is established. In order to prevent the problems of the particle swarm algorithm from falling into local minima and slow convergence, the particle swarm algorithm is improved by using a strategy combining inertia weight non-linear decreasing and adaptive learning factor, so as to improve the training accuracy of the model.Taking the measured GPS data of a mining area as an example, we verify the fitting performance of the model. The results show that the improved wavelet neural network model has higher accuracy and stability in GPS height fitting.