An adaptive OD flow clustering method to identify heterogeneous urban mobility trends

Abstract

Origin-Destination (OD) flow, as an abstract representation of the object's movement or interaction, has been used to reveal the movement patterns of human activities and the coupling process of the human-land system. As a developing spatial analysis method, OD flow clustering can be used to identify the dominant trends and spatial structures of urban mobility. However, urban flow exhibits universal heterogeneity, which is mainly manifested in irregular shapes, uneven distribution, and obvious scale differences. The existing methods are constrained by specific spatial scales and sensitive parameter settings, making it difficult to reveal heterogeneous urban mobility patterns within travel OD data. In this paper, we propose an OD flow analysis method that integrates spatial statistics and density clustering. This method can determine parameter values from datasets without manual intervention and adaptively identify multi-scale mixed OD flow clusters. In the simulation experiment, the proposed method accurately detects all preset OD clusters with less noise. It outperforms the baseline methods in terms of Silhouette Coefficient, V-measure, and Fowlkes Mallows index. As a case study, this method is applied to OD data from Chengdu, China, extracting 63 representative flow clusters and revealing the trends of heterogeneous urban mobility across different lengths and densities for public transit optimization.