Unraveling the dynamics of urban park visits: Daytime visitation patterns and behavioral determinants derived from mobile phone data

Time-of-day patterns of urban park visits are essential for evaluating user engagement and health benefits. While park usage patterns across the day are shaped by both the timing and duration of visits, these two dimensions are rarely examined together at a citywide, year-round scale. This study delved into the time-of-day patterns and behavioral determinants of park visits using mobile phone data and machine learning techniques. Leveraging a one-year Tokyo mobile phone dataset provided by ZENRIN DataCom Co., LTD, we plotted time-of-day fluctuation curves (TFCs) for ten comprehensive parks across eight representative days (weekdays and weekends in each of the four seasons) and applied a multilabel classification model build on XGBoost to identify factors influencing visitors' timing and duration choices. The results showed that the intensity of park usage varied across the day, ranging from high midday-afternoon peaks on spring weekends to low daytime activity on winter weekdays. Our model achieved strong predictive accuracy (55.92% for timing, 60.51% for duration) and revealed that contextual factors such as weather and travel distance were more influential than static park features in shaping individual visit decisions. The findings introduced a temporal dimension to urban park planning, allowing for dynamic, demand-responsive design strategies that align park features with usage patterns and evolving visitor needs.