Inferring nonwork travel semantics and revealing the nonlinear relationships with the community built environment

The relationship between the built environment and nonwork travels reveals potential supply-demand status between urban functional facilities and diverse living travels, which provide critical indicative information for urban spatial structure optimization. Benefiting from the large-scale travel datasets, it would be possible to investigate deeper research by modeling the implied nonwork travel semantics (such as travel purposes). This study designs a framework to reveal the nonlinear relationships between the community built environment and the potential semantics of nonwork travel trajectories. First, we construct indicators to characterize the built environment of individuals’ living communities. Then, focusing on the definition of activity areas around drop-off points, latent Dirichlet allocation (LDA) is utilized to infer the implied semantics of nonwork travel. The random forest model is employed to quantify the relationships between the community built environment and nonwork travel with various semantics. The experimental results obtained with real-life datasets from Haikou, China uncover both compensatory nonwork travel derived from diverse facility accessibility and the contributions of housing attributes to travel semantic discrepancies. The findings provide scientific references for the optimization of urban spatial layouts of both functional facilities and residential communities to meet the living demands of diverse residents without long-distance intracity travel.