Abstract: Through field survey and data collection, 18 influencing factors involving landform, geology, hydrometeorology, human activities, rock-soil mass properties and vegetation coverage are selected to evaluate the landslide vulnerability in Jiyuan City of Henan Province on the basis of information-convolutional neural network (CNN) coupling model, with GIS spatial analysis to quantify the spatial distribution characteristics of landslide. The results show that the landslide disasters are distributed aggregately in the area, with multiple high value centers of kernel density. The areas with very low, low, medium, high and very high landslide susceptibility account for 45.04%, 34.58%, 8.67%, 9.12% and 2.57%, respectively. The very high and high susceptible areas are characterized by developed faults, fragile geological environment and hydraulic erosion. The highest landslide density is 0.804 per km², occurring in medium susceptible area. The ROC curve and AUC value indicate that the evaluation results have good accuracy, and the prediction ability of the coupling model is reliable. The top 5 influencing factors of landslide susceptibility analysis are distance from roads, distance from faults, slope aspect, terrain roughness, as well as erosion degree and type. This study may provide scientific basis for the prediction and prevention of landslide geological disasters in cities and towns on the Loess Plateau.