川西地区地形复杂多变，地质构造活跃，致使交通干线的建设与维护面临着地质灾害频发的挑战。集成学习算法能优化地质灾害易发性评估中算法的不足，提升模型的精度，在地质灾害易发性评估中具有显著优势。以沿江高速为例，选取坡度、起伏度等12个特征变量构建地质灾害易发性评价体系，比较分析Stacking集成算法与单一算法建模的预测性能，探讨了沿江高速地质灾害的主控因素，并验证了模型的实用性。结果表明：沿江高速沿线地质灾害高和极高易发区占比分别为18.21%和9.85%，集中分布在雷波段和金阳段；Stacking集成模型受试者操作特征（receiver operating characteristics, ROC）曲线和精确率-召回率（precision-recall，P-R）曲线的曲线下的面积（area under curve, AUC）均达到0.89，F₁分数也达到0.84，显著高于3个单一机器学习模型的ROC曲线的AUC（0.84~0.86）、P-R曲线的AUC（0.81~0.85）和F₁分数（0.78~0.79），比单一机器学习算法有更好的预测性能；沿江高速地质灾害的发育受地形地貌因素控制；新增灾害点位于模型中的极高易发区，印证了Stacking模型的精确度和可靠性。

The terrain in western Sichuan is complex and varied, and the geological structure is active, which makes the construction and maintenance of the traffic trunk line face the challenge of frequent geological disasters. Ensemble learning algorithm can optimize the shortcomings of the algorithm in geological hazard susceptibility assessment and improve the accuracy of the model, which has significant advantages in geological hazard susceptibility assessment. Taking the riverside high-speed as an example, 12 feature variables such as slope and relief were selected to construct the geological hazard susceptibility evaluation system. The forecasting performance of the modeling of the integrated algorithm and a single algorithm was compared and analyzed. The main control factors of the geological disasters along the riverside high-speed were discussed and the practicability of the model was verified. The results show that the proportion of high and extremely high geological hazard prone areas along the Yangtze River high speed is 18.21% and 9.85%, respectively, which are concentrated in the Leibo section and Jinyang section. The area under curve (AUC) of the receiver operating characteristics (ROC) curve and the precision-recall (P-R) curve in the integrated model. The AUC of ROC curve (0.84~0.86), the AUC of P-R curve (0.81 ~ 0.85) and the F₁ score (0.78 ~ 0.79) of the three single machine learning models are significantly higher, and the prediction performance is better than that of a single machine learning algorithm. The development of high-speed geological hazards along the Yangtze River is controlled by topographic and geomorphic factors. The new damage points are located in the highly prone areas of the model, which verifies the accuracy and reliability of the Stacking model.