Dynamic community detection and evolution analysis in multilayer interaction network based on trajectory data

Dynamic community with life cycle is a collection of geographical entities in spatial interaction networks. They connect closely through human movement compared with other spatial entities. However, there is still a lack of systematic approaches for analyzing spatiotemporal variation. In this regard, this paper first constructs the spatio- temporal interaction multilayer network (STIMN) based on human mobility data. Secondly, we extract dynamic communities of STIMN, which reveal the uneven distribution of urban resources and the regularity and diversity of human movement. Thirdly, based on the theories of landscape ecology, we propose a dynamic community evolution model and analyze the clustering patterns, forming the framework of "the community detection - dynamic evolution - clustering analysis". Finally, we apply the framework in the third- ring core area within Wuhan, which shows that: (1) Communities extracted from public activities have similarities and differences with the existing administrative divisions of Wuhan City, which can better reflect the people-oriented urban dynamic spatial organization; (2) Community evolution shows an apparent life cycle of "occurrence - expansion - stability - contraction - disappearance"; (3) The life cycle of different communities is different. Through clustering, communities can be divided into short-term, medium- term and long- term communities. Each type of community has one or more dynamic patterns, such as stability, saddle- shaped and wave- shaped pattern, which are significant to dynamic urban planning and management. This study breaks the limitations of traditional community analysis methods, helps explore the dynamic characteristics of the STIMN and deepens the understanding of the dynamic man-land interaction. © 2023 Science Press. All rights reserved.