A prediction model for Xiangyang Neolithic sites based on a random forest algorithm

Abstract The archaeological site prediction model can accurately identify areas to enable better knowledge and understanding of human civilization processes social development patterns. A total 129 Neolithic data in the region were collected using Xiangyang area as study area. An eight-factor index system elevation, slope, slope direction, micromorphology, distance water, position, planar curvature, profile curvature was constructed. geospatial database with a resolution 30 m × established. whole sample set built trained ratio 1:1 nonarchaeological sites obtain results. average Gini coefficient used evaluate influence various factors. results revealed that under curve values receiver operating characteristic curves 1.000, 0.994, 0.867 for training, complete, test datasets, respectively. Moreover, 60% historical, located high-probability zone, accounting 12% proposed this matched spatial distribution characteristics locations. With assessed best samples, categorized into three classes: low, average, high. proportion low-, average-, zones decreased order. mainly near second third tributaries distributed at low eastern hills central hillocks. random forest (RF) rank importance variables. Elevation, micro-geomorphology classified most important RF has stability predictive ability case field; provides new research method reference revealing relationship between activities natural environment.