Abstract: The aim of this research is to improve the accuracy of predicting landslide deformation. Firstly, trend terms and error terms were isolated through denoising the deformation data by using Kalman filter. Then comprehensive graph weight method was employed to determine the combinatorial weights for trend terms. Furthermore, neural network model was adopted to the prediction for error terms. Results suggest that the effect of half parameters and half Kalman filter method is the optimum. The present model has improved prediction accuracy, and is verified to be of feasibility.

Key words: landslide, deformation prediction, regression model, neural network, comprehensive graphic weight method