Smart Cities, Journal Year: 2025, Volume and Issue: 8(2), P. 58 - 58
Published: March 30, 2025
Understanding and predicting urban vitality—the intensity diversity of human activities in spaces—is crucial for sustainable development. However, existing studies often rely on discrete sampling points single metrics, limiting their ability to capture the continuous spatial distribution vibrancy. This study introduces UVPN (urban vitality prediction network), a novel deep-learning architecture designed generate high-resolution predictions static dynamic at regional scales. The integrates two key innovations: SE (squeeze-and-excitation) block adaptive feature recalibration an RCA (residual connection with coordinate attention) bottleneck position-aware learning. Applied New York City, leverages diverse morphological features such as streetscape attributes land use patterns predict distributions. model outperforms architectures, achieving reductions 34.03% 38.66% mean squared error population density pedestrian flow predictions, respectively. Feature importance analysis reveals that road networks predominantly influence density, while strongly affect flows, built interest contributing both dimensions. By advancing prediction, provides robust framework evidence-based planning, supporting creation more sustainable, functional, livable cities.
Language: Английский