Monitoring and predicting reservoir slope deformation, which is influenced by water level fluctuations and rainfall infiltration, is essential for slope stability assessment and early warning. Long short-term memory (LSTM) networks capture long-term dependency features in time-series data, while random forest (RF) models address complex nonlinear relationships among multiple variables. A method for slope deformation prediction, integrating LSTM and RF, was proposed. Small baseline subset-interferometric synthetic aperture radar (SBAS-InSAR) technology was used to process Sentinel-1 imagery from 2018 to 2022, generating a time-series deformation field for the mainstream bank slopes in the middle-western section of the Three Gorges Reservoir Area. The reliability of the InSAR inversion results was validated with global navigation satellite system (GNSS) monitoring data. Precipitation and water level data, along with deformation data, were coupled as inputs for the prediction model. The annual average deformation rate in the study area ranges from -59.31 to 36.4 mm/year, with acceleration during periods of concentrated rainfall and water level fluctuations. Predictions show that cumulative deformation will peak at -177.2 mm between June and September 2022. The root mean squared error (RMSE) of the test set ranges from 3.410 to 3.968 mm, and the mean absolute error (MAE) ranges from 2.522 to 3.088 mm. The model exhibits good prediction accuracy and stability, confirming the applicability of the method for predicting reservoir bank slope deformation.