Dynamic station criticality assessment of urban metro networks considering predictive passenger flow

Metro systems serve as vital lifelines in high-density cities. Identifying critical stations within the metro systems is a crucial aspect of metro network construction and management, playing a vital role in ensuring the functionality and sustainability of the urban transit networks. Previous research has indicated that the criticality of metro stations is influenced by both their network properties and the fluctuations in passenger flow. However, comprehensive studies that accurately capture and predict the dynamic fluctuations in metro station criticality are limited. In response, this paper introduces a predictive and comprehensive station criticality quantification (PCSCQ) approach that integrates predictive passenger flow data into the decision-making model to assess and forecast metro station criticality. This method synthesizes complex network theory (CNT), deep learning, and multi-attribute decision-making (MADM) into a cohesive framework. In this framework, CNT analyzes the metro network properties and provides station centrality data; deep learning processes the large-scale metro passenger flow data and generates accurate future flow predictions; and MADM consolidates various metrics to produce final quantifications and predictions of station criticality. The Hong Kong Mass Transit Railway (MTR) system is used as a case study to verify the feasibility and effectiveness of the proposed approach. The results indicate that the proposed method effectively captures fluctuations in station criticality. For instance, Kowloon Tong Station is identified as the most critical station in the Hong Kong metro network during most of the day on a typical weekday, while Mong Kok Station and Tsim Sha Tsui Station become more critical during the evening peak hours. Moreover, by incorporating predictive data, the proposed method can forecast changes in station criticality, particularly in scenarios where stations experience unexpected incidents and lack historical data for reference. This research contributes to system planning, resource allocation, and emergency preparedness in metro systems.