Abstract: This paper attempts to accurately evaluate the stability of debris slopes located in the eastern Tibet section of the G318 highway. It can provide a reference for the highway public sector to carry out maintenance, with a view to thoroughly solve the serious damage of debris slopes to this section. We investigated all the debris slopes along the section of the district 1277km in the past 3 years. According to the composition of the debris, the size of the particles and the characteristics of the movement, the basic characteristics and failure mode of all kinds of debris slopes are described. The correlation between the main factors and the slope stability is analyzed with the information entropy and FLAC numerical simulation. The stability evaluation model of the debris slope based on support vector machine is established using the analysis results of the influence factors. The results show that:(1)Debris slopes can be subdivided into 3 types of sliding sand slopes, rolling rock slopes and gravel slopes. The characteristics of the 3 types of debris slopes are as follows:The sand slipping is of small cale, but happens suddenly and often in groups; The rolling stone slope is mainly composed of spalling or partially slipping of alluvial and diluvial deposits, with a large number and great harm. The gravel slopes are sudden burst with large energy, high speed, a certain scale and great harm. (2)Information entropy theory is used to analyze the correlation degree of 10 factors on the stability of the debris slope. The effective factors in turn are rainfall snow melting, average slope, debris, average slope height, vegetation coverage, engineering construction, earthquake, weathering degree, ground shaking and river erosion. (3)Numerical simulation shows the main influencing factors. The slope safety coefficients are compared after normalization with its curve slope relationship:rainfall snow melting(0.8) > average slope(0.66) > average slope height(0.52) > vegetation coverage(0.46) > construction(0.43). The greater the gradient of the curve, the greater the correlation between the factor and the stability of the debris slope. The conclusion is consistent with the field investigation and analysis. (4)Combining the information entropy and the results of numerical simulation using FLAC analysis, six of the main factors are selected for the stability evaluation model of the debris slope. The model accuracy rate is 89.29% and 93.75%respectively through the model training, back estimation and test, which can be used for the stability evaluation of debris slope to give some advises and reference for maintenance.