Nonlinear effects of multilevel factors on public transport commuting in China’s cities

The demand for urban public transport is closely associated with the built environment, as analysed at both the city and neighbourhood levels. Although neighbourhood-level built environments have been studied thoroughly, city-level factors have received less attention. In this study, a machine learning framework is developed to explore the influence of multilevel factors on public transport commuting. A case study of 79 cities in mainland China reveals: (1) The highest accuracy was achieved by incorporating individual socioeconomic and commuting characteristics, and neighbourhood- and city-level built environments. (2) The cumulative importance of built environment factors exceeds 50%, with approximately half contributed by city-level variables. (3) All level factors exhibit nonlinear effects, with some variables displaying threshold effects. (4) Cities are classified into five distinct groups based on city-level variable effects. These findings highlight the necessity of incorporating multi-level factors into planning practices and targeted policy recommendations for different types of cities.