Abstract：Aiming at the difficulty of accurate prediction of seismic observation data, we propose a regression model based on nuclear mixed effects. In order to verify the feasibility of the algorithm model, we perform a simulation experiment with the output data of the geophysical instrument of the Hubei Seismic Station and compare it with the traditional neural network algorithm. The results show that the model can accurately predict the seismic and geophysical observation data and the performance is better than other neural network algorithms. The relative error of the water temperature and water level data prediction is less than 0.05% and 0.48%. The proposed model provides a new research idea for earthquake monitoring and forecasting personnel to accumulate and analyze basic earthquake data. At the same time, it also provides practical foundation and research possibilities for more complex deep learning algorithm framework models.
