Abstract: Landslide displacement prediction is an important part of monitoring and warning work. Machine learning algorithm has high prediction accuracy and is widely used in landslide displacement prediction,but its interpretability is poor. For this reason,this paper proposes an interpretable landslide displacement prediction model(XGB-SHAP)that combines the Extreme Gradient Boosting(XGBoost) algorithm and the Shapley Additive Expansions(SHAP). Taking the Duoying landslide in Ya ′an,Sichuan Province as an example,the landslide displacement time curve was decomposed into trend term and periodic term,and factors such as previous rainfall and previous displacement were selected as input features of the model. Through XGBoost and ARIMA(Autoregressive Integrated Moving Average model),the displacement of its cycle term and trend term are predicted respectively,and finally the prediction model is explained. The results show that the mean absolute error and mean square error of XGB-SHAP are 2.46 and 3.38 respectively,which is more accurate than other models. The results show that the mean absolute error and mean square error of XGB-SHAP are 2.46 and 3.38 respectively,which is more accurate than other models. The results of model interpretation show that the displacement increases obviously when the 30-day cumulative displacement exceeds 20 mm or the 7-day maximum rainfall exceeds 19 mm. The model interpretation results show that the displacement increases significantly when the cumulative displacement in the current period exceeds 20 mm in 30 days or the maximum rainfall in 7 days exceeds 19 mm.