Abstract: Dingjie County in Tibet is located in a high-altitude region characterized by complex geological conditions, active tectonics, and frequent geological hazards. Conducting a regional susceptibility assessment of geological hazards is therefore critical for effective disaster prevention and mitigation. Focusing on Dingjie County, this study establishes a susceptibility evaluation framework based on ten influencing factors: elevation, slope gradient, lithology, terrain relief, distance to fault zones, distance to rivers, distance to roads, mean annual NDVI, mean annual precipitation, and settlement kernel density. These factors form the basis for constructing a geological hazard susceptibility evaluation index system. First, the geographic detector method was applied to quantify the explanatory power of each factor with respect to hazard susceptibility. To address the limit availability of hazard inventory data, a generative adversarial network (GAN) was employed to augment training samples. Subsequently, a GAN-enhanced Random Forest (GAN-RF) model was developed to produce a five-class susceptibility map: low susceptibility, relatively low susceptibility, moderate susceptibility, relatively high susceptibility, and high susceptibility. Finally, model validation and SHAP visualization were conducted to assess model performance. Geodetector results indicate that elevation, mean annual precipitation, and settlement kernel density are the dominant controlling factors. Model validation shows that while the traditional Random Forest (RF) model achieved an AUC of 0.897, the GAN-RF model significantly improved this to 0.953, markedly enhancing the accuracy of identifying high-susceptibility zones. Notably, the high-risk zones, covering only 1.40% of the county area, concentrated 85.81% of the verified geological hazard points. The GAN-RF model effectively improves susceptibility mapping accuracy in high-altitude regions, particularly optimizing the identification of high-risk zones. SHAP validation confirms its effectiveness in evaluating such areas, providing reliable basis for disaster prevention and mitigation decisions.