Abstract: Huanglongxicun landslide is located in Tianshui city of Gansu province. It belongs to loess landslide with high potential energy. With the volume capacity of 3.9105m3, its substratum is granodiorite. In order to improve the prediction accuracy of grey GM(1,1) model, center approach grey GM (1,1) model is used in the deformation prediction of Huanglongxicun landslide. It is indicated that the predicting values of center approach GM (1,1) model are well agreed with monitoring values for the landslide, and have much higher prediction accuracy than that of traditional grey GM(1,1) model because error sum of squares and error percent for center approach GM (1,1) model is lower than that of traditional grey GM(1,1) model. Moreover, the center approach GM (1,1) model can be improved by adjusting the value of the parameter m in the model. When m is equal to 6, prediction accuracy for center approach GM (1,1) model is over 5.34% than that of traditional grey GM(1,1) model.