Neighbourhood Type, Built Environment and Bus Use in Beijing : A Machine-Learning Analysis

The relationship between urban form and travel behaviour in Chinese cities has received considerable interest. However, it remains unclear how built environment may exert differentiated effects on people’s transit use across neighbourhood types that reflect the planning and development trajectories of Chinese cities. Focusing on Beijing, this study first identifies typical neighbourhood types characterised by distinctive development features. Next, by employing smart card data coupled with a machine-learning approach (XGBoost), we identify that regional accessibility exhibit stronger association with bus use for transient and suburban neighbourhoods, whilst local attributes appear to be more influential for inner-city neighbourhoods. Further, the effects of built environment factors including density and availability of various facilities on trip frequency and distance are found to vary markedly across neighbourhood types in terms of effective range and thresholds. As such, planning strategies to reshape travel behaviour shall better consider the relevance of neighbourhood type