Establishing a high-precision displacement prediction model is of great significance for landslide early warning. However, accurately predicting the complex nonlinear growth of landslides remains a challenging task. In this paper, we present a groundbreaking model for forecasting mountain landslide displacement. The proposed approach incorporates ensemble empirical mode decomposition (EEMD) to efficiently decompose landslide cumulative displacement into two distinct components: trend and periodic. By employing a least squares polynomial function, the trend displacement is accurately predicted, while the periodic displacement is forecasted using a convolutional neural network-long short-term memory (CNN-LSTM) model. The final cumulative displacement prediction is obtained by combining the predicted trend displacement components and the periodic displacement components through a superposition procedure. Taking the typical Shuping landslide in the Three Gorges Reservoir area as an example, the experimental results demonstrate the feasibility of the EEMD-CNN-LSTM model. The model achieves a root mean square error (RMSE) of 11.18 mm and a coefficient of determination (R²) of 0.96, indicating high prediction accuracy. It effectively extracts data features, reduces interference information, and compensates for the limitations observed in static prediction models. This provides a useful approach for landslide displacement prediction in mountainous regions.