Abstract: Prediction model in landslide displacement is the key part for building landslide early warning system. The reliability and accuracy of prediction model mainly depend on the main controlling factors and the basic theoretical model. Researchers have already made a great achievement on the displacement prediction models according to practical cases. However, insufficient understanding due to multi-factors influence on landslide movement still exist, because of the strong individual feature and complex tendency forecasting in landslide movement. Kualiangzi landslide, a typical type of translational landslides in Southwest of China, is selected in this paper to make a deep research based on previous data collection. Grey relational analysis and correlation coefficient analysis are used to ensure the main controlling factors influencing landslide movement. A model combining GM(1, 1) grey model and the wavelet neural network optimization(WNN)model optimized by the improved adaptive genetic algorithm(IAGA) is presented. Through mining and analyzing the monitoring data of Kualiangzi landslide for five years, the results show that landslide movement is influenced by the accumulative rainfall, osmotic pressure, groundwater table and soil moisture content. The predicted results are in good agreement with the real-time monitoring data. After comparative analysis, the results show that in terms of the stability and accuracy, this model are better than the models of traditional BP neural network, wavelet neural network and GA-WNN. The new model has a good application prospects in the field of landslide early warning and forecasting.