Data-driven predictive modeling of user stay behavior from audiovisual simulation for metro commercial space design

Despite increasing recognition of the commercial potential of metro systems, a design-oriented framework for predicting how sensory comfort influences stay behavior—a key determinant of commercial viability—remains underdeveloped. While multisensory effects on shopping behavior have been widely studied, efficient quantitative tools linking sensory comfort to users’ stay behavior are still scarce, limiting evidence-based design of metro commercial spaces. On-site observations indicate that the audiovisual environment plays a pivotal role in shaping behavioral responses, whereas other sensory factors (e.g., thermal comfort) exhibit limited variability across metro commercial settings, making their behavioral effects less readily distinguishable despite their potential implicit influence. Therefore, this study develops a data-driven model to quantitatively evaluate the effects of the audiovisual sensory environment on the stay behavior quality in metro commercial spaces. Field studies and virtual reality experiments were integrated to collect real-time data on audiovisual conditions and behavioral responses across diverse metro commercial spaces. Artificial neural networks and regression analysis were applied to investigate relationships between audiovisual conditions and stay behavior. The results indicate that: (1) audiovisual comfort significantly influences stay behavior, with visual comfort explaining up to 60 % of behavioral variance, compared to 50 % for acoustic comfort; (2) spatial pattern serves as a critical moderator in the sensory–behavior relationship. To translate these empirical findings into practice, a pre-occupancy evaluation model is proposed to support data-driven design decisions. Thus, a practical framework is offered for the design of metro commercial spaces that enhance audiovisual comfort and maximize commercial potential.