<scp>CFlow</scp> : A Novel High‐Order Flow Analysis Unit for Detecting Spatiotemporal Mobility Patterns Based on Origin–Destination Data

ABSTRACT Mobility behavior research has long been a focal point in geographic information science (GIS). Many researchers use isolated OD pairs as flow analysis units when studying mobility behavior based on OD (origin–destination) data. However, isolated OD pairs only reflect individual mobility, which may constrain applying a collective analytical perspective, hampering the exploration of the potential value of OD data. In this research, we investigate combining isolated OD pairs in close spatiotemporal proximity to form aggregated edges and introduce a novel high‐order flow analysis unit, collective mobility flow (CFlow), for mining spatiotemporal mobility patterns. Compared to traditional OD research, CFlow provides richer contextual information by weaving together spatiotemporal proximity OD pairs. Such transformation has leveraged the sequence‐based data structure that empowers OD analysis with more intelligent tools. Furthermore, this paper designs relevant indicators for CFlows to analyze its spatiotemporal characteristics, including time span and area coverage. In particular, we apply classical frequent pattern mining algorithms to CFlows to explore frequently occurring collective spatiotemporal mobility patterns. To evaluate the effectiveness of the proposed CFlows extraction and analysis framework, extensive experiments have been carried out using OD data from cruise taxis and online taxis in Xiamen, China. Experiment results reveal spatiotemporal variations in CFlows under the two taxi operational modes, “taxis seeking passengers” and “passengers seeking taxis”, validating the feasibility of the proposed methods. The proposed CFlow analysis methods enable the discovery of spatiotemporal mobility patterns in long‐term spatiotemporal interactions, thus enriching the modeling toolset for a comprehensive understanding of spatiotemporal interactions in OD data.