Landslides are one of the most destructive geological hazards worldwide, posing serious impacts on roads, human safety, and the ecological environment. Therefore, susceptibility assessment plays a crucial role in scientifically and systematically preventing and mitigating landslide disasters. Taking the Chengdu Panda National Park as the study area, a hybrid model was developed based on the VIKOR method, frequency ratio (FR) method, and three classification algorithms: Random Forest (RF), CatBoost (CB), and LightGBM (LG). A total of 20 triggering factors, including topographic, hydrological, and soil-vegetation factors, were selected. The spatial correlation between factors and landslide occurrence was studied by using the frequency ratio, and the importance of factors was calculated using the three algorithms. Based on this, a susceptibility map for landslides in the region was created using the hybrid model. The accuracy and precision of the models were evaluated using the Receiver Operating Characteristic (ROC) curve, Area Under the Curve (AUC), and Kappa coefficient. The results showed that the VIKOR-FR-LG model performed the best in predicting landslide susceptibility. Elevation, proximity to rivers, and distance from fault lines were found to be significant factors affecting landslide occurrence, and the area with extremely high sensitivity covered 2473.30km2. Therefore, VIKOR-FR-LG is the most effective model for predicting landslide susceptibility in the study area, and the research findings can provide reference for landslide prevention and ecological safety management in the Chengdu area of the Giant Panda National Park.