Abstract: In order to accurately predict the deformation trends of landslides and prevent disasters in advance, a landslide deformation trend prediction model based on VMD-SSA-LSTM was proposed.Firstly, the Variational Mode Decomposition (VMD) model was used to decompose the landslide GNSS monitoring curve into multiple sub-signals.Then the Long Short-Term Memory network (LSTM) optimized by the Sparrow Search Algorithm(SSA) was employed to analyze the trend prediction of each sub-signal.Finally, the predicted values were summed to obtain the trend prediction results.By comparing this model with LSTM, VMD-LSTM, and BP neural network models, the prediction performance was verified.The results show that VMD-SSA-LSTM prediction model had a higher adaptability of complex temporal features, as well as higher training and testing accuracy compared to other models.The prediction results were consistent with the actual deformation results, demonstrating its high adaptability and feasibility for landslide trend prediction analysis.