Abstract: The Kezhayi to Gongnaisi section of the Duku expressway is predominantly characterized by alpine landforms, with steep terrain cutting that provides conducive conditions for avalanche development. The study on the evaluation of snow avalanche susceptibility in this area is a crucial prerequisite for the safety construction and operation of the Duku expressway. The 149 snow avalanche points were collected by employing remote sensing interpretation and field investigations. Through correlation analysis of these factors, 10 evaluation factors were selected, forming the avalanche evaluation factor system. Subsequently, the non-avalanche points and original avalanche points were extracted using the K-means clustering method and random method to create a sample set. Machine learning techniques, including multilayer perceptron (MLP) and support vector machine (SVM) algorithms, were utilized to assess avalanche susceptibility in the study area. The results show that the sample datasets extracted by the random and K-means clustering methods were used for training, the Kappa coefficient of the R-SVM, R-MLP, K-SVM, and K-MLP models were greater than 0.6. These four sets of models exhibited a high degree of consistency between the predicted results and actual values of the validation dataset. The AUC (area under curve) value trained by MLP increased from 0.762 to 0.983, while the AUC value trained by SVM increased from 0.724 to 0.951. Based on the K-MLP model partition with the highest evaluation accuracy, the snow avalanche development in the research area has a relatively minor impact on the proposed route but may pose a threat to tunnel entrances. This study provides theoretical support and methodological references for the construction, operation and mitigation of sonw avalanche disasters for the Duku expressway.