Heterogeneous intensity-based DBSCAN (iDBSCAN) model for urban attention distribution in digital twin cities

Abstract

Urban regions confront a wide array of safety and management challenges, including natural disasters and the pressing demand for real-time incidents response. Information technology supports to collect massive multi-dimension data to form a digital twin-based urban system. However, as urban environments become increasingly complex and dynamic, traditional monitoring systems struggle to deliver timely and comprehensive alerts for urban management. To address these limitations, this paper formulates a heterogeneous intensity-based DBSCAN (iDBSCAN) clustering model for dynamic urban monitoring under digital twin environment. The proposed iDBSCAN integrates crucial factors such as traffic flow, population density, and urban activities to dynamically segment urban regions into clusters with differentiated attention levels. By leveraging spatiotemporal data, iDBSCAN enables adaptive allocation of computational resources, responding dynamically to changing urban conditions and the severity of events. To further refine the clustering analytics, this research incorporates Alpha-Shape for outlining the geometric boundaries of attention regions and Kernel Density Estimation (KDE) for smoothing the regional attention distribution, providing a more nuanced understanding of urban structures and dynamics. A computational experiment is conducted based on the spatial–temporal urban data in Haikou, which validates the effectiveness of iDBSCAN in identifying clusters with varying attention levels and capturing the intricate dynamics of urban environments.