Abstract: It is of great significance to establish the evaluation model and evaluate geological disaster susceptibility for improving the efficiency and accuracy of regional geological disaster forecasts and early warning. However, how to establish a geological disaster susceptibility evaluation model which is suitable for regional reality and has the value of popularization and application is the key scientific problem that restricts the prediction and early warning of geological disasters. Taking Nanhua County in Yunnan Province as the research example, based on the detailed survey data of geological disasters in 2015, 11 factors were selected, including distances from roads, distances from rivers, distances from faults, soil types, precipitation, land use types, rock mass types, vegetation coverage, slopes, aspects, and elevations. Based on the mean method, the evaluation study on the susceptibility of geological disasters in Nanhua County was carried out by using the gradient boosting tree algorithm (XGBoost, LightGBM, CatBoost), the information model, and the geographically weighted regression model. The results showed that: ① The prediction results of the geographically weighted regression model had the phenomenon of over-fitting, and the information model had the phenomenon of under-fitting. ② The mean method had the best effect. The AUC value was 0.9337, and the accuracy was respectively 1.7%, 1.8%, 2.0%, 3.8%, and 4.0% higher than that of a geographically weighted regression model, XGBoost, LightGBM, CatBoost, and information volume model. ③ Catboost had the worst prediction effect on positive samples, but the highest prediction effect on negative samples; XGBoost had the best prediction effect on positive samples and poor prediction effect on negative samples; while the mean method based on three gradient algorithms had significantly improved the prediction of positive and negative samples. ④ The main inducements of geological disasters in Nanhua County were road constructions, fault activities, rainfall rushes, and river erosion. The high susceptibility areas were characterized by their closeness to rivers, roads and faults.