Passenger data complexity in tram stop dwell time modelling

Abstract

The stop dwell time can be modelled by using the volumes of boarders and alighters, and it is a common conclusion that the use of additional information on the number and width of doors, number of seats, and number of through standees in model creation improves its estimation of stop dwell time. However, such an approach demands detailed knowledge and/or assumptions on passenger distribution both inside the vehicle and on the stop platform, which makes the model creation and its application more challenging. The research presented in this paper is focused on the passenger input data requirements for the creation of tram stop dwell time prediction models. It is based on passenger and tram dwell time data collected at an island tram stop in Zagreb. The data acquisition included the field recording of the trams in operation during five working days, laboratory processing of 70 hours of collected video data, and creation of a synthesized database of observed and measured data. Three different multiple linear regression models for tram dwell time prediction were created, with the following independent variables: (1) the volume of boarders and alighters and a type of passenger flow transiting through the busiest tram doors, (2) the volume of boarders and alighters transiting through the busiest tram doors, and (3) the total volume of boarders and alighters per tram. The cross-validation of the model showed that passenger input data simplification has a minor effect on the model’s goodness of fit, and a mild effect on it’s accuracy and precision, which could be adequately addressed by the application of a larger operating margin.