Service disruptions significantly alter train congestion patterns, reducing the usefulness of standard real-time congestion information for passenger decision-making. To address this issue, this study proposes a model for predicting train congestion levels during disruptions. The model was evaluated using actual disruption data and achieved approximately 75% accuracy. The results indicate that features such as headway and congestion levels of the same train at one to three stations prior to the prediction target station are effective for congestion prediction during disruptions.
UEDA et al. (Fri,) studied this question.