Taking Baishuihe landslide in the Three Gorges Reservoir Area as an example, the features of landslide deformation influencing factors such as rainfall, reservoir water and groundwater were extracted using kernel principal component analysis method, which were used to build the BP neural network displacement prediction model and then to output the fitted values. Secondly, Markov chain (MC) was used to optimize the prediction error through analysis of the error range between fitted values and measured values, timing displacement state division and state transfer probability matrix calculation. Finally, the BP-MC model was constructed, using which dynamically displacement prediction was achieved. The fitted and predicted results show that this model can reflect the relationship between the inducing factors and landslide displacement and can effectively improve the prediction accuracy of landslide displacement.