A review of train delay prediction approaches

Railway operations are vulnerable to delays. Accurate predictions of train arrival and departure delays improve the passenger service quality essential for real-time railway traffic management minimise their further spreading. This review provides a synoptic overview discussion covering breadth diverse approaches predict We first categorise research contributions based on underlying modelling paradigm (data-driven event-driven) mathematical model. then distinguish between very short long-term classify different input data sources that have been considered in literature. discuss advantages disadvantages producing deterministic versus stochastic predictions, applicability during disruptions interpretability. By comparing results included contributions, we can indicate prediction error generally increases when broadening horizon. find data-driven might edge event-driven terms accuracy, whereas explicitly model dynamics dependencies strength providing interpretable more robust concerning disruption scenarios. The growing availability is expected increase appeal big-data machine learning methods. • provide an extensive literature methods delay prediction. Methods classified model, data, other criteria. strengths weaknesses types derive insights horizon, quality, used sources. identify gaps trends, underscore future directions.