A Multi-source Data Fusion Approach for Pre-opening Passenger Flow Prediction at New Metro Stations

This study addresses the operational needs during the initial phase of new metro line openings by proposing a framework for time-segmented passenger flow prediction at new stations. Using AFC data from Beijing and Nanjing as empirical samples, intra-day consistency is evaluated via the coefficient of variation, while intra-week differences are measured using Euclidean distance. A two-tier “shape–scale” clustering approach based on an improved K-Means++ algorithm is adopted to construct a spatiotemporal passenger flow profiling library. Built environment indicators such as population, employment, and land use are integrated, and cross-classification combined with stepwise regression is employed to establish correspondences between profiles and environments. A “round-by-round classification” procedure ultimately yields time-segmented forecasts for new stations. Empirical validation demonstrates that this method achieves higher accuracy and interpretability than conventional approaches, providing practical support for capacity allocation and passenger service organization in the early operation stage.