The prediction of passenger flow under transport disturbance using accumulated passenger data

Abstract

Listen

Translate article

Whenever train disturbances occur, it is necessary for traffic operators to recover the train timetable appropriately, considering the passenger flow. But it is difficult to predict the flow in quantitative terms, because passengers may cancel his or her travel, or detour to another rail line. In recent years, however, it has become possible to obtain actual train operation time data stored in traffic control systems, the number of passengers on board by means of load compensating devices on rolling stock and passengers’ Origin-Destination data collected with automatic ticket checkers. In this paper, the authors first propose a visualization method of passengers’ flow. The method makes it easier for the authors to understand features of passengers’ flow during traffic disturbances in comparison with that of ordinary days. In the next step, the authors construct prediction models for the number of passengers passing the section between two adjacent stations. The authors implement multiple regression analysis using passenger’s flow data and information on outline of disturbances on a commuter rail line in past 10 months. As a result, the authors get multiple regression formulas to predict increase or decrease rates of the traffic volume in each section, with sufficient multiple correlation coefficients about 0.75. Finally, the authors apply the formulas to other disturbances, and find that they are reliable enough to support train rescheduling operations.