A Geographic Information Technologies and Deep Learning Approach to Exploring the Mismatch Between Street Vitality and Convenience

Keywords: Street vitality, Convenience, Spatial heterogeneity, Deep Learning
Abstract Type: Paper Abstract

Abstract

This study addressed the gap in assessing street vitality in high-density urban environments, focusing on Sham Shui Po's central district in Hong Kong. Data incompatibility and scale limited traditional methods using demographic or economic metrics, particularly in walkable cities. Geographic information technologies and machine learning enabled more efficient and comprehensive assessments of street vitality.
The study developed a novel framework for assessing street vitality, grounded in Maslow's hierarchy of human needs. Previous studies have shown that convenience is a critical variable in street vitality. The deep learning model was used to simulate the degree to which pedestrians perceived the vitality of the street. Geospatial analysis was used to identify mismatches between pedestrian vitality perception and convenience, with Sham Shui Po as a case study.
The geographically weighted regression model highlighted spatial variation in convenience's impact on vitality, revealing the spatial heterogeneity of vitality factors across different areas. In areas like Lai Chi Kok Road and Nam Cheong Street, oversaturation of amenities negatively affected vitality. In contrast, areas like Nam Cheong and Tai Nam Streets, lacking amenities, saw positive impacts from added facilities. These insights suggested that policymakers should focus on improving existing facilities in oversaturated areas and expanding amenities in underserved regions.
This research advanced the understanding of street vitality in high-density urban settings, providing a replicable framework for targeted urban interventions through human-centered theories and geospatial technologies.