Detecting Community Structure of Urban Hotspot Regions

The travel behavior of residents is influenced by environment factor such as urban transportation system and administrative division. In turn, users equipped with navigation devices act as sensors detecting the environmental dynamics. The long-term accumulation of navigation big data contains massive valuable spatio-temporal information. We propose to detect the spatio-temporal distribution and community structure of urban hotspot regions from navigation big data. A framework including data preprocessing, hotspot region detection, urban spatial discretization, and community structure detection is designed in this work. Hotspot regions are detected by kernel density estimation and density-based clustering on origin-destination (OD) points of navigation trajectories. The hotspot regions are discretized into Voronoi polygon grids based on the spatial distribution of OD points. Finally, we analyze the complex network formed by hotspot region grids and employ Louvain algorithm to detect the community structure of hotspot region network. This framework is implemented on the taxi dataset of Chengdu. The experimental results reveal the spatio-temporal distribution and community structure of urban hotspot regions in different periods of weekdays and weekends. The urban hotspot regions and the community structure are influenced by inherent geographical environment and dynamically evolve with time. The spatio-temporal characteristics of urban hotspot regions are proved to be the result of coaction of environment and human activities. Findings of this work could provide decision-making support for transportation system optimization, city layout plan, and smart city construction etc.