Quantifying MyCiTi supply usage using Big Data and Agent-Based Modelling

Abstract

Cape Town's Bus Rapid Transit (BRT) system, MyCiTi, uses an Automated Fare Collection (AFC) system that generates large volumes of transactional data on a daily basis. This data can be considered Big Data. The AFC data in its raw format, however, is incapable of supporting supply and demand analysis (e.g. studying bus occupancy rates). Agent-Based Modelling (ABM) can be used to analyse such data for that purpose. This paper discusses the development and calibration of a MATSim-based ABM to analyse AFC data for Cape Town's BRT system. It is shown that data-formatting algorithms are critical in the preparation of data for modelling activities. Furthermore, the development of appropriate ABM calibration parameters requires careful consideration in terms of appropriate data collection, simulation testing, and justification, which are discussed. The paper furthermore shows that the calibrated ABM can generate outputs such as bus on-board volumes, a system-demand overview, and even individual commuter path choice behaviour. Finally, a validation exercise shows that the model developed for this study is able to provide good estimates of on-board bus volumes (R2 = 0.85). It is, however, recommended that further research be conducted into studying agent path choices through simulation.