Opportunity with a threshold: a SHAP-based machine learning explanation of key drivers and their nonlinear relationships with rural investment

Rural investment is a crucial trigger for mitigating rural decline. While existing research has identified multiple determinants of rural investment, systematic comparisons of their relative importance remain scarce, and potential nonlinear dynamics remain underexplored. This study integrates machine learning models with SHapley Additive exPlanations (SHAP) to identify key drivers of rural investment across 689 villages, focusing on socio-economic factors such as new immigrants and indigenous returnees, as well as measures of structural embeddedness, including cuisine and surname centrality. These factors exhibit inverted U-shaped relationships, indicating that excessive embeddedness may constrain capital inflows. This interpretable framework elucidates rural investment patterns and offers guidance for balancing sustainability trade-offs in rural revitalization.