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

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