Determination of peak times in public transportation

Abstract

Big data and advances in data mining promise to overcome major challenges in public transportation that result in improvements in network efficiency and rider satisfaction. In this study, we develop methods i) to determine peak and off-peak hours and ii) to cluster boarding profiles of demand throughout a day for all bus lines of the public transportation network. With the outcome of this study, it will be possible to propose bus scheduling algorithms that take into account the dynamic nature of ridership patterns by both time and location. As a case study, the passenger boarding patterns of the Antalya transportation network were analyzed. Among the bus lines, 2 are selected to demonstrate that the peak and off-peak times are not necessarily the same, and a dominant demand peak or several demand peaks are possible. Based on the peak demand characteristics, buses may be scheduled.