Machine Learning Maps How Ancient Hongshan Settlements Rose and Reorganized

A new study of the Hongshan culture in northeastern China traces how prehistoric settlements in the Liaoxi region shifted over time, combining archaeological records, paleoclimatic data, and geospatial analysis with interpretable machine learning.

The research team analyzed 1,190 recorded sites, of which 67 have well-defined chronological phases, allowing them to follow settlement dynamics across three distinct periods.

Site numbers followed a clear trajectory of 6, then 37, then 24 across the three phases, a rise and partial decline that coincided with the emergence of large ceremonial centers during the culture’s development.

Environmental factors played a persistent role in where people settled. Proximity to rivers and the aspect of slopes consistently influenced site placement throughout all periods, underscoring the importance of water access and terrain orientation in prehistoric location choices.

Social drivers, by contrast, changed markedly over time. In the middle period, craft production dominated the factors shaping settlement patterns, while in the late period, ritual-oriented functions became the leading influence.

To disentangle these factors, the researchers applied SHAP analysis, an interpretable machine learning technique that quantifies each variable’s contribution to model predictions, making the drivers of site distribution transparent rather than hidden inside a black-box model.

The SHAP results pointed to a systemic reorganization of social networks, with interactions shifting from positive to negative between phases, suggesting changing relationships among communities as the culture evolved.

Hongshan culture is considered a critical case for understanding prehistoric social complexity in northeastern China, and the new framework links spatial analysis directly with archaeological interpretation.

The authors describe their approach as a replicable methodology for studying prehistoric settlement patterns, one that could be applied to other regions where site catalogs and environmental data are available.

Because the findings rest on recorded site inventories, with only 67 of the 1,190 sites carrying secure chronological assignments, the phased conclusions apply most firmly to that well-dated subset, and the study is published in npj Heritage Science.

Subject of Research: Spatiotemporal evolution and influencing factors of Hongshan cultural settlement sites in the Liaoxi region, Northeast China

Article Title: Spatiotemporal evolution and influencing factors of Hongshan cultural sites in the Liaoxi Region, Northeast China

Article References: Wu, R., Shang, J., Chang, Z., & Li, Y. (2026). Spatiotemporal evolution and influencing factors of Hongshan cultural sites in the Liaoxi Region, Northeast China. npj Heritage Science. https://doi.org/10.1038/s40494-026-03057-7

Image Credits: AI Generated

DOI: 10.1038/s40494-026-03057-7

Keywords: Hongshan culture, Liaoxi region, Northeast China, settlement patterns, spatial analysis, machine learning, SHAP analysis, archaeology, prehistoric social complexity, paleoclimate, ceremonial centers, geospatial data