Modeling vibrant areas at nighttime: A machine learning-based analytical framework for urban regeneration

Abstract

Enhancing vibrant areas at nighttime (VAN) is important for promoting urban regeneration. However, the process simulation of the potential impact of urban renewal initiatives on VANs has yet to achieve complete coupling with population mobility, economy, and land utilization. In this study, we developed a simulation framework to simulate the changes in VANs and reveal their potential interconnection with urban regeneration strategies. The research methods involved the use of overlaying multiple data sources, including satellite imagery, thermal and land use maps, and point of interest data, to obtain a collection of nightlife activities in Mianyang City. We were able to identify and grade VANs based on multisource big data, providing support for urban renewal planning, as well as determine the key driving factors and configuration patterns of environmental elements that impact VAN. We also developed a machine learning-based predictive model for urban regeneration and VAN redevelopment. These results show that a vibrant nightlife can be used to regenerate dilapidated urban areas, thus reducing urbanization. Moreover, the simulation method developed in this study has wide applicability in other regions for identifying potential improvements and guiding investment and revitalization efforts in a targeted and effective manner.