Abstract: In order to accurately grasp the deformation and development laws of landslides in the accumulation lager on the bank of the Three Gorges Reservoir area, the authors conducted a comprehensive study on landslide deformation trends using the rescaled range method, grey model and optimized generalized regression neural network, based on the landslide deformation monitoring data. The research results show that the Hurst exponent of each monitoring point is greater than 0.5 in the landslide deformation trend discrimination results, indicating a continuous increasing trend of landslide deformation. In the deformation prediction results, with the continuous optimization and combination processing of the GM (1,1)-SFLA GRNN model, the prediction accuracy has been significantly improved, indicating that the model construction process is reasonable. Besides, its prediction shows that landslide deformation will continue to increase. The average relative error of the obtained prediction results is between 1.76% and 1.82%, and the training time is between 52.21 ms and 57.23 ms, which has a better prediction effect. Then, BP neural network and support vector machine were introduced for analogical prediction. The results show that the GM (1,1)-SFRA-GRNN model has relatively higher prediction accuracy and faster training speed compared to BP neural network and support vector machine. Comparing the results of landslide deformation trend discrimination and the results of deformation prediction, the authors found that landslide deformation will continue to increase without convergence trend. The necessity of landslide prevention and control is significant, and the rationality of the two analysis methods is mutually supported, providing certain theoretical support for landslide prevention and control.