Transit-oriented Development (TOD) is promoted to foster sustainable urban development by enhancing the integration between transportation and land use. To evaluate the complex relationship between rail transit TOD construction and urban development, an evaluation index system is established, taking into account land use, degree of mixing, station accessibility, station carrying capacity, and station service capacity. By utilizing machine learning methods, specifically Random Forest, along with road network data and other relevant sources, the impact of TOD on urban development in Harbin, a city with a cold climate, is analyzed. The study reveals that conventional bus routes, station location, and road network density have a significant influence on urban development. Furthermore, the SHAP model emphasizes the threshold effect associated with TOD construction and the interplay among TOD variables. These findings provide valuable insights for planning and revitalization efforts in public transportation-oriented urban development.