Abstract: The common slopes classification systems are suitable for a road, railway or a certain area landslide risk assessment, which aims to ordering the priority of the landslide count-measurements. The study of using slope classifying system to assess the landslide risk is rare in open literature. The environmental conditions of loess collapse are similar, whereas the controlling factors to the collapse have differences at individual slopes. So the scoring and grading are a good way to assess the risk of the loess collapse. This paper selects a loess ravine area of 3.2 km2 in suburban district of Yanan, Shannxi, China as study area. It is implemented on the basis of analysis to the assessing factors and their weights, which have been done in previous works. It uses the hydrology methods to zone the assessing units, sets four grading standard for the hazards factors and five grading standard for the consequence factors, and synthesizes the hazards results and the consequence results to set the risk judge matrix. Accordingly, it designs the slope classifying system for the loess collapse risk assessment. Applying this classification system at 37 loess slope units, the result shows that the collapse risk level of each unit has not a regular distribution. It only depends on the local geological conditions and the elements in the risk. The statistics of units risk shows phenomena of "much in middle and less in both ends", which means the area of "high risk" and "very low risk" is very less (only 27%), whereas the area of "moderate risk" and "low risk" is much more (73%). This classifying system can be used as a semi-quantitative method to assess the collapse risk at the loess slopes.