Abstract: Xiangxiang city, a typical county-level city located in the central part of Hunan province, was taken as the research object, and the ecological security evaluation index system was constructed from 14 indexes, such as elevation, slope, topographic relief, land use type, NDVI, distance to river, distance to highway, rainfall, biodiversity, soil thickness, grain yield, etc. An ecological security evaluation model based on random forest （RF） intelligent algorithm was proposed, which combined with GIS technology to evaluate the ecological security of 50 m×50 m grid was used as the evaluation unit to identify the extremely important factors affecting the ecological security level of Xiangxiang city. The results showed that the ecological security level of Xiangxiang city was in the middle and low level, and the proportion of low and medium level grid was 84.33%. The spatial distribution of ecological security level was uneven, which was higher in the west and north than in the east and south, higher in the hilly area than in the plain, and higher in the mountainous area than in the urban area. The main factors affecting the level of ecological security in Xiangxiang city were biodiversity, distance from highway, distance from river and land use type in order of importance. Soil thickness, soil erodibility, rocky desertification degree and other factors had relatively little influence on the evaluation of ecological security level. The sample training accuracy of the random forest ecological security evaluation model was 98.01%, and the test accuracy was 95.53%, which had a high accuracy.

Key words: ecological security evaluation, random forest model, Xiangxiang city