Spatial differentiation and driving mechanisms of traditional villages based on geo-explainable artificial intelligence.

Traditional villages represent important cultural landscapes in China; however, their spatial patterns and driving mechanisms remain unclear in the context of rapid urbanization. This study selected 194 traditional villages and 4,705 spatial grids within the Ganjiang River Basin as the research sample. By integrating kernel density analysis, spatial autocorrelation methods, and a geo-explainable artificial intelligence (GeoXAI) framework, this study systematically examined the spatial distribution characteristics and influencing factors of traditional villages. The results revealed that traditional villages exhibit a highly clustered spatial pattern, with a Global Moran's I of approximately 0.985, forming a distinct "core-edge" structure. Sunshine duration, elevation, and urbanization rate were identified as the dominant explanatory variables, all of which demonstrated nonlinear effects and spatial heterogeneity. Given the potential for reverse causality and scale mismatch, these associations are interpreted as correlational rather than causal. The GeoXAI model significantly enhances the interpretability of spatial mechanisms and underscores the importance of integrated geospatial variables. The findings indicate that the spatial differentiation of traditional villages is the outcome of multifactor interactions involving natural geography, socioeconomic development, and settlement structures. This study contributes empirical evidence useful for understanding the spatial evolution of traditional villages and offers data-driven guidance for heritage conservation and spatial planning in Jiangxi Province and other regions.