Quantifying the cascading effects of passenger delays

Delays in transport networks has adverse implications for infrastructure and service managers as well as travellers. While it is widely acknowledged that delays occurring across the transport networks may be related, there are is lack of knowledge on the underlying properties of these relations and means for quantifying them. To this end, we develop a network-wide data-driven delay analysis method. First, we construct a Bayesian network to represent the relations between delays associated with different transport network elements and assess the reliability of critical infrastructure elements. Second, we propose a series of original metrics denominated informativity indicators for quantifying the spatial extent of the delays observed based on the Bayesian network obtained. The proposed approach is applied to the Washington DC metro network. Time-dependent passenger waiting and transferring delays inferred over more than a year from smartcard data are utilized as input to the Bayesian network. Our findings indicate that passenger delays at few selected stations are directly informative of delays occurring at many other stations. We also examine the relation between the proposed informativity metrics and the topological properties of metro stations, concluding that the latter have a limited value in approximating network-wide delay correlations.