Abstract: While urban rail transit systems alleviate urban issues and optimize spatial layout, they also reshape housing price distributions. Existing research, however, rarely explores the heterogeneous impact mechanisms across various station-area types or identifies the nonlinear threshold effects of influencing factors. Focusing on Hangzhou, China in 2025, this study integrates multi-source open data to construct a Node-Place-Environment (N-P-E) model and employs the K-Means algorithm to classify 192 metro station areas into five distinct types. The nonlinear relationships among housing price determinants are unveiled by integrating Shapley Additive Explanations (SHAP) and Partial Dependence Plot (PDP) methods with the Extreme Gradient Boosting (XGBoost) model. Results reveal that: 1) station areas exhibit a core-periphery spatial differentiation pattern. The transit-oriented (TOT) and balanced and coordinated (BCT) types dominate the urban core, whereas the comprehensive service-oriented (CST), basic supporting-oriented (BST), and underdeveloped (UDT) types cluster in the periphery. 2) The core driving factors of housing prices are inherently associated with the functional positioning of station areas, confirming the logic of function-demand adaptation. 3) The impacts of the built environment exhibit nonlinearity and threshold effects that differ across station-area types. An insufficient or excessive supply of key elements triggers negative externalities, such as congestion and Not In My Back Yard (NIMBY) effects. This study provides a scientific basis for optimizing residential resource allocation along rail transit lines and formulating differentiated station-area planning strategies.