Application of link prediction algorithms in urban rail transit networks

Accurate link prediction in urban rail transit networks (URTNs) presents unique challenges due to their fixed spatial constraints and operational complexity. Current methodologies for abstract networks fail to address these structural specificities, particularly the interdependencies between stations and lines. This study introduces an innovative bipartite graph embedding framework that integrates topological constraints with passenger flow dynamics to overcome these limitations.Our approach features three key advancements: (1) A heterogeneous graph decomposition strategy preserving line-station relationships through bipartite mapping, enabling explicit modeling of transfer dependencies; (2) An enhanced Bipartite Network Embedding (BINE) algorithm incorporating queuing theory-driven passenger flow simulations, where node sampling frequencies adapt to station service capacities; (3) A multi-criteria similarity optimization model combining topological and operational features through advanced weighting mechanisms. Experimental validation across multiple metropolitan networks demonstrates significant improvements in network resilience and operational efficiency. The framework effectively balances topological integrity with passenger flow optimization, showing enhanced robustness against structural disruptions and improved transfer efficiency. These outcomes provide actionable insights for phased network expansion strategies, establishing a new paradigm for transit network optimization that bridges theoretical graph modeling with practical urban planning requirements.