Spatio-Temporal Analysis of Urban Leisure Consumption Activities Based on Autoencoder and Multi-Source Data: A Case Study of Chongqing, China

Urban leisure consumption activities (ULCA) are key indicators of urban vitality, and analyzing their spatio-temporal distribution is essential for understanding the structure and dynamics of urban spaces. This study conducts a quantitative analysis of leisure consumption behavior in the main urban area of Chongqing based on multi-source data (including mobile signaling data, POI data, and traffic analysis zone) and proposes a novel spatio-temporal layout analysis method. The study focuses on the spatial layout of ULCA venues, the spatio-temporal distribution of activity behaviors, and their interrelationships. An autoencoder model is employed to predict behavioral patterns and visualize leisure consumption dynamics. Specifically, five key indicators—land diversity, population density, transportation accessibility, environmental livability, and site convenience—are defined to quantify and evaluate urban leisure consumption vitality. Subsequently, spatio-temporal trends of leisure consumption are analyzed using the autoencoder, and trajectory predictions of human behavior are performed to reveal vitality patterns across different areas of Chongqing. Finally, cluster analysis and tag cloud visualizations are used to identify ULCA hotspots and explore their spatial characteristics. The results indicate significant spatio-temporal variation in ULCA distribution, reflecting disparities in urban vitality across regions. This study provides valuable data support for urban planning and leisure facility design, while also offering new perspectives and methodological insights for future urban vitality research.