Objective This research aims to quantify the geocultural characteristics of Chinese traditional villages through spatial network analysis, principal component analysis (PCA), and clustering algorithms. By constructing a national-scale agglomeration demarcation model, this research proposes differentiated conservation strategies to optimize conservation area delineation methods considering natural geography, historical culture, and socioeconomic dimensions. The research ultimately provides theoretical foundations for policy-making to balance heritage preservation with regional development dynamics.

Methods The research method adopted in this research is to combine graph theory and machine learning with data from the lists of six batches of traditional villages published by the Ministry of Housing and Urban-Rural Development of the People’s Republic of China and other departments. First, the minimum Spanning tree (MST) model weighted by the actual transportation time cost is adopted to carry out the preliminary clustering. This process quantifies the spatial connection strength between villages by prioritizing the selection of low-cost connections and cutting off high-cost connections, and ultimately forms a spatial structure that highlights transportation accessibility and spatial proximity. Then, through Principal Component analysis (PCA) and K-Means clustering methods, the multi-dimensional cultural and geographical characteristics are analyzed. The analysis objects not only include the clustering results obtained based on the MST method, but also cover natural and cultural elements. On this basis, this research constructs a characteristic spectrum based on the geographical, economic and cultural characteristics of traditional villages, and divides the villages into different agglomeration areas according to their natural geographical attributes and cultural attributes. And through the spatial superposition analysis method, the K-Means clustering method are observed to explore the cultural background, language system characteristics, environmental characteristics and architectural types involved in each zoning.

Results The MST and K-Means clustering results are respectively spatially superimposed and analyzed with the current situation of provincial administrative divisions across the country. The results show that the MST clustering boundaries based on transportation cost weighting are highly coexisting with the provincial administrative boundaries in provinces with dense traditional villages. This phenomenon stems from the characteristics of the MST method. The network construction principle of MST, which aims to minimize the overall transportation cost, makes it easier to form continuous clusters overlapping with administrative boundaries in areas with high density of traditional villages and strong transportation accessibility. Administrative boundaries are often delineated relying on geographical barriers such as rivers and mountains, and these barriers can lead to a significant increase in transportation costs, which is corresponding to the segmentation logic of MST that removes high-cost edges. In contrast, the K-Means clustering boundaries based on the MST clustering results and combined with the main components of natural and cultural elements such as topography, mountain and river aggregation, climate, and language system have a relatively high consistency with the provincial administrative boundaries in provinces with scattered traditional villages. The distribution of traditional villages in these provinces shows the characteristics of "broad geographical space" and overall “great dispersion”. The geographical and economic and cultural characteristics extracted by the PCA method show a spatial polarization of “small clusters”. In provinces with a high concentration of traditional villages, the delineated clusters are mostly distributed at the junctions of provincial administrative boundaries. This regularity indicates that the transportation network correlation and segmentation involved in MST can serve as the basis for the initial division of traditional village clusters. The division result of the “MST + PCA + K-Means” model can rely on the extraction ability of PCA for geographical, cultural and other characteristics, and form a comprehensive division result on the basis of simultaneously explaining the intrinsic connection between the geography and culture of each region.

Conclusion A comparative analysis of the MST clustering model weighted by actual transportation time cost, as well as geo-cultural zoning boundaries and administrative boundaries shows that when the MST clustering boundary is highly consistent with the transportation cost gradient, and the historical rationality of traditional zoning lies in its implicit consideration of transportation cost efficiency. A significant deviation indicates regionalization may be dominated by non-transport factors like cultural association and resource dependence; especially when the MST boundary crosses the transportation cost gradient, further analysis of cross-gradient cultural or economic drivers is needed. The MST weighted by transportation cost and the “principal component + K-Means” model essentially reflect the methodological opposition between “spatial determinism” and “cultural reductionism”. For research objects like traditional villages featuring the interaction of multiple scales and multiple influencing factors, it is necessary to break through the bottleneck of single quantitative analysis, comprehensively consider the combination of geography, culture and economy, and allow coexistence of transportation networks, language diffusion and other relationships to approach the true self-organization of characteristic traditional village clusters.