Nonlinear Effects of Traffic Supply and Land Use on Employment–Residential Ratio in Transit-Oriented Station Areas: Evidence from the Highest-Density Built-Up Zone of Shenzhen

The worsening jobs–housing imbalance in high-density urban areas has become increasingly prominent. However, existing studies predominantly focus on the city and regional scales, with limited attention to small-scale spatial units, failing to provide empirical evidence for refined regulation of transport and land use. As a typical compact development mode at the small scale, Transit-Oriented Development (TOD) offers an ideal spatial unit for examining micro-level jobs–housing relationships. However, its influence on jobs–housing balance remains unclear. This study takes 72 built subway station areas in Shenzhen’s highest-density built-up area as the research object. Employing quadratic polynomial regression and XGBoost-SHAP methods, it reveals the effects of transportation supply and land use on the employment–residential ratio, as well as threshold effects and indicator synergies. The main findings are as follows: (1) Transportation supply (Node), land use (Place), and TOD degree (TODness) all exhibit U-shaped effects on the employment–residential ratio, with inflection points at Node = 0.330, Place = 0.412, and TODness = 0.776, identifying the critical transition from balance to employment polarization. (2) Number of subway directions is the most critical indicator, together with betweenness centrality, floor area ratio, population size, and land use entropy, each exhibiting notable threshold effects. (3) A significant interaction effect exists between betweenness centrality and land use entropy, suggesting that traffic network connectivity and land use evenness can jointly regulate the employment–residential ratio in both directions. This study provides a quantitative basis for phased regulation of jobs–housing relationships in high-density TOD station areas, with findings transferable to other high-density cities.