Dynamic patterns of intercity mobility and influencing factors: Insights from similarities in spatial time-series

Substantial spatiotemporal variations are rooted in intercity mobility due to diverse characteristics of urban systems, especially during holidays, yet remain poorly understood at present. This research aims to bridge this gap by presenting a novel dynamic perspective on intercity mobility and its influencing factors. Specifically, we captured general time-varying patterns of intercity mobility while considering spatial differences and exploring temporal similarity. To achieve this, large-scale spatial time-series of intercity mobility flow across both non-holiday and holiday periods were extracted from the nationwide Location-based Services (LBS) dataset in China. The research framework is divided into three phases: Firstly, we employed shape-based clustering to capture the general time-varying patterns by exploring their similarities. Afterward, we identified crucial features that determine time-varying patterns using Extreme Gradient Boosting (XGBoost), incorporating a comprehensive feature set related to urban attributes and spatial connection. Lastly, we provided an explainer that specifies the feature contributions using the SHapley Additive exPlanations (SHAP). Our findings revealed similarities in dynamic patterns of intercity mobility flow, suggesting that different cities or city pairs potentially exhibit similar time-varying trends. Also, we noted significant diversities in dynamic patterns across various periods, primarily characterized by peak flow trends around weekends and the start/end of holidays. Feature analysis identified population density and land use intensity as crucial factors shaping mobility flow patterns of cities, while distance and economic connection largely influenced mobility flow patterns between city pairs. Comparison results indicated that the same features exert differentiated effects and varying intensities across different periods.