Uncovering the spatiotemporal motif patterns in urban mobility networks by non-negative tensor decomposition

Abstract

Investigating the macroscopic mobility laws of the population and the microscopic travel characteristics of individuals in a city offers an essential way of understanding the city as a complex system. Research that combines the travel laws of the population and individuals simultaneously poses a challenging and complicated task. In this paper, a 4-dimensional tensor is created to describe the spatiotemporal characteristics of people’s various mobility motifs. Non-negative tensor decomposition is used to identify the principal patterns of time, space, and daily motifs in a city and the level of interactions between them. Moreover, the major network indicators of the human mobility networks constructed in terms of three principal motif patterns are statistically analyzed and calculated from the complex network perspective. Relying on a conjoint analysis of core tensor and network indicators, we find that the three motif patterns correspond to three real-life mobility scenarios: simple motifs dominated by commuting, complex motifs based on more complicated life and entertainment activities, and single trips departing from or arriving at airports or railway stations, respectively. Furthermore, the intra-city mobility networks constituted based on the three motif patterns differ significantly in network heterogeneity, node importance, entropy, and clustering coefficient. This suggests that the existing studies on intra-city human mobility networks that only consider first-order trips may be the average results of aggregations containing many people with different characteristics. In our work, the specific research on the spatiotemporal characteristics of people with different mobility motif patterns can assist policy makers in conducting fine-scale management and implementing specific policies.