Passenger to Train Assignment Using Only Smart Card Data

Abstract

With the introduction of smart card systems, sophisticated data collection in urban railways has become possible. One key challenge is identifying the trains taken by passengers, which relies on synchronizing smart card data with train arrival data. This paper proposes a train-level assignment method based solely on smart card data. The proposed approach consists of four steps: tap-out time clustering, train arrival generation, train schedule generation, and trip assignment. First, tap-out times from smart card data were used to cluster trip-alighting patterns using the mean shift algorithm. Second, the train arrival schedule was generated by labeling each group with the earliest tap-out time. Third, these station-specific arrival schedules are connected to generate the overall train routes based on the passengers’ travel times between origins and destinations. Lastly, trips were assigned to the generated train schedules, and they were used to estimate congestion levels. The proposed approach was applied to Seoul metro Line 9 in South Korea. The results showed that the generated schedule was consistent with the actual train arrival data and produced coherent operational patterns across all stations. With the generated train schedules, the trip assignment was conducted, and the results showed that 48.7% of passengers used local trains, while 51.3% used express trains. The congestion levels were also identified with the generated train schedule and assigned trips. As such, the proposed approach contributes to trip assignment using smart card data in a simple manner.