<p>The landslide susceptibility assessment (LSA), which presents the probability of a landslide occurrence under site-specific geological and environmental conditions, is essential to prevent and mitigate the landslide risk. The machine learning is one of important techniques for the study on LSA, and various models about the machine learning differ from each other in the LSA. In this paper, the Meishan township, Zhejiang Province, China was selected as a case study, and the Pearson correlation assessment and multicollinearity test were conducted to measure the relationship among 12 factors such as elevation, slope, and slope aspect. The training models were separately established with the decision tree (DT), random forest (RF), gradient-boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and convolutional neural network (CNN), and the results were assessed by means of the area under the curve (AUC) of receiver operating characteristic plot as well as the frequency ratio (FR). The results demonstrated that the CNN achieved the best performance of AUC=0.874, followed by XGB (AUC=0.806), GBDT (AUC=0.801), FR (AUC=0.794), and DT (AUC=0.735), indicating that the DT-based ensemble models predict better than the single DT models. Furthermore, the CNN model, due to its complex structures, performs better than conventional machine learning model in dealing with the nonlinear problems.</p>