A Bayesian network modeling for departure time choice: A case study of beijing subway

Abstract

Departure time choice is critical for subway passengers to avoid congestion during morning peak hours. In this study, we propose a Bayesian network (BN) model to capture departure time choice based on data learning. Factors such as travel time saving, crowding, subway fare, and departure time change are considered in this model. K2 algorithm is then employed to learn the BN structure, and maximum likelihood estimation (MLE) is adopted to estimate model parameters, according to the data obtained by a stated preference (SP) survey. A real-world case study of Beijing subway is illustrated, which proves that the proposed model has higher prediction accuracy than typical discrete choice models. Another key finding indicates that subway fare discount higher than 20% will motivate some passengers to depart 15 to 20 minutes earlier and release the pressure of crowding during morning peak hours.