The predication of landslide deformation is of great importance in landslide control and engineering construction. An Improved Variable Dimension Fractal-Artificial Neural Network (IVDF-ANN) coupling model is proposed for landslide monitoring. According to the characteristics of slope evolution, landslide is found to be a nonlinear dynamic system whose monitoring data are fractal. The model is established based on improved variable dimension fractal to predict the trend of time series in association with artificial neural network to optimize the deviation of time series. As a case study, the model is applied to the prediction of the displacement of Maoping Landslide with its in-situ displacement data. The result shows the model gives full display to the self-similarity of fractal theory and the self-learning ability of artificial neural network, and thus brings about vast range of prospect for application due to its high precision and noise immunity.