Abstract: The Zagunao River Basin in Lixian County features a typical deep-cut river canyon landscape characterized by active geological processes and frequent landslides. This study integrates the geological conditions of Lixian County with the deformation and developmental characteristics of landslides in the Zagunao River Basin. Sixteen influencing factors, including elevation and terrain cutting depth, were selected across three categories: topography, geology, and natural factors. The frequency ratio model was employed to quantify the relationship between these factors and landslide occurrences. Landslide susceptibility assessments were then conducted using Random Forest(RF) and eXtreme Gradient Boosting(XGBoost) models at both grid cell and slope unit levels. Model performance was evaluated using ROC curves and confusion matrices. Additionally, the deformation rate maps of the evaluation units were superimposed to achieve a precise susceptibility assessment of the Zagunao River Basin. The results indicate that terrain cutting depth, stratigraphy, and land use were the most significant factors among topography, geology, and natural factors. Both RF and XGBoost models showed similar importance rankings for influencing factors. In grid cell assessments, stratigraphy emerged as the dominant factor, whereas terrain cutting depth was the primary controlling factor in slope unit assessments. ROC curve evaluations revealed AUC values of 0.942, 0.928, 0.934, and 0.920 for RF-grid cells, RF-slope units, XGBoost-grid cells, and XGBoost-slope units, respectively. The confusion matrix results demonstrated that the RF-grid cell model outperformed the other models, with recall, precision, and accuracy scores of 0.935, 0.956, and 0.945, respectively. These findings highlight the RF-grid cell model as the most accurate approach for assessing landslide susceptibility in the Zagunao River Basin, providing valuable insights for disaster prevention and mitigation efforts in Lixian County.