Abstract: To carry out the geological hazard susceptibility zoning and prevention in Harbin, Heilongjiang Province, eight evaluation factors including slope gradient, slope aspect, curvature, lithology, NDVI, distance from river, distance from road and distance from structure are selected to establish the evaluation index system of geological hazard susceptibility. The non-geological hazard samples are randomly chosen from the extremely low and low susceptible zones calculated by information algorithm, which forms the document data set together with the geological hazard samples. Besides, three ensemble learning methods such as random forest(RF), Adaboost and Stacking are used to assess the geological hazard vulnerability in Harbin, and the accuracy is verified by confusion matrix. The results show that the trend of evaluation results of the four algorithms is the same, and consistent with the actual situation of the study area. The major inducing factor of geological hazards in Harbin is human engineering activities, with the extremely high susceptible zones mainly concentrated near roads. The area of extremely high susceptible zones predicted by RF algorithm accounts for only 1.27% of the whole region, yet the number of geological hazards takes up 21.03%, with the frequency ratio of 16.58 and the maximum AUC value reaching 0.891, indicating that RF algorithm has more advantages in the geological hazard susceptibility evaluation among the above three algorithms.