Considering the characteristics of nonlinearity and complexity of concrete dam deformation monitoring data, in order to improve the accuracy of concrete dam deformation prediction, a concrete dam deformation prediction model based on the improved empirical modal decomposition (EMD) method and the long short-term memory (LSTM) neural network was proposed. This model adopts the wavelet threshold denoising method to optimize the high-frequency components decomposed by the EMD method, effectively removing the data noise while retaining the characteristic information of the original data as much as possible. The LSTM neural network was used to perform time series prediction on the processed data. The results of case validations show that this model can accurately simulate the deformation process of the dam body, demonstrating a high prediction accuracy.