Behavior‐Oriented Inference of Trip Purposes and Urban Mobility Structures From Dockless Bike‐Sharing Data

Dockless bike‐sharing systems generate large volumes of passively collected mobility data, creating new opportunities for understanding urban activity patterns and travel behavior. However, inferring trip purposes from dockless bike‐sharing data remains challenging due to the absence of explicit activity labels and the complexity of mixed‐use urban environments. Using dockless bike‐sharing trip records from 11 sampled days in 2021 in Shenzhen, China, we develop a behavior‐oriented framework for inferring trip purposes by integrating gravity models with Bayesian rules, functional land use, temporal preference, and nighttime light intensity. The inferred results reveal highly interpretable spatiotemporal mobility structures, including pronounced weekday commuting rhythms, mixed‐use intra‐category circulation, and strong functional interactions among home, work, schooling, and transfer‐related activities. To further examine mixed‐use urban mobility behaviors, a semantic adjustment strategy and post hoc activity‐chain exploration are introduced, providing alternative interpretations of mixed‐purpose movements and plausible record‐linked sequential activity patterns embedded within cycling trips. Our results further demonstrate strong temporal consistency and high spatial replicability under machine learning approximation, indicating that the inferred behavioral structures capture stable and reproducible spatiotemporal regularities that can be learned across dates. These findings show that even under incomplete and weakly supervised mobility data conditions, dockless bike‐sharing trips contain reproducible behavioral structures that can support interpretable urban activity analysis, providing new opportunities for behavior‐oriented mobility analysis in data‐scarce urban environments.