Dynamical community detection and spatiotemporal analysis in multilayer spatial interaction networks using trajectory data

Abstract

Detecting network communities has recently attracted extensive studies in many fields. However, little attention has been paid to detection and analysis of dynamical communities. This study intends to propose a methodological framework to detect dynamical communities in multilayer spatial interaction networks and examine their spatiotemporal patterns. Random walks are used to merge network layers with different weights, the Leiden technique is used for deriving dynamical communities and exploratory analytic methods are adopted to examine spatiotemporal patterns. To verify our methods, experiments were conducted in Wuhan, China, where trajectory data were used to construct the time-dependent multilayer networks. (1) We derived a set of spatiotemporally cohesive and comparable dynamical communities on each day for one week; (2) They exhibit interesting clustering patterns according to the similarity of their growth curves; (3) They display distinct life courses of occurrence, expansion, stability, contract and disappearance, and their dynamical interactions are vividly depicted; (4) They manifest mixed land use patterns via transfers of human activities. Thus, our methods can enrich research on dynamical organization of urban space and may be applicable in other contexts, while experimental results can provide decision-making support for sustainable urban management.