Abstract: Loess has unique water sensitivity and structural properties, and is extremely sensitive to water. In addition, extreme rainfall events on the Loess Plateau have significantly increased in recent years, and regional shallow loess landslides have occurred frequently, making this region one of the most frequent landslide prone areas in the world. This study takes the shallow loess landslide watershed in Chengguan District, Lanzhou City, Gansu Province as the research object. Through field investigation, data collection, and indoor mechanical experiments, data on influencing factors and model parameters were obtained. Combined with XG Boost machine learning model and Scoops3D physical mechanics model, the risk analysis of shallow loess landslides under different rainfall conditions was carried out. The results show that the extremely high and high prone areas are mainly located in the mountainous and hilly areas on both sides of the Lanzhou Basin, accounting for 16.42% and 3.75% of the total area, respectively. The top four factors contributing to landslide impact are slope, rainfall, elevation, and terrain undulation; The stability analysis results of the landslide show that with the increase of rainfall intensity, light rain conditions have the most significant impact on the increase of the area of the unstable and unstable zones, and the overall slope stability in the region is developing towards an unstable and unstable state. The hazard analysis results of 贴the landslide show that the influence area of the extremely dangerous area in the study area changes most obviously under light rain, heavy rain and rainstorm conditions, with an area increase of at least 1.35km². The aim of this study is to conduct a risk assessment of rainfall induced shallow loess landslides using a combined model, providing effective scientific basis for the development of human settlements and the safe operation of major engineering projects in the Chinese Loess Plateau region.