Validating T-BEST Models with 100% APC Counts


Xuehao Chu

Document Type


Publication Date


Subject Area

infrastructure - stop, infrastructure - vehicle, infrastructure - interchange/transfer, planning - route design, land use - planning, mode - bus


Accessibility, Automated passenger counters, Boarding, Boarding equations, Bus routes, Bus service, Buses, Improvements, Peak periods, Public transit, Service planning, Simulation, Stopping, Transfer boarding, Vehicle fleets


The current version of the Transit Boarding Estimation and Simulation Tool (T-BEST) uses a set of boarding equations that were estimated with a small sample of boarding data. In addition, it uses an overall weekday peak that combines the morning and afternoon peaks. This report documents a research effort in re-estimating a set of boarding equations for T-BEST to achieve two objectives. One objective was to use boarding data that are far more reliable than what were used in estimating the current set of boarding equations. The other objective was to make improvements in the boarding equations. Both objectives have been successfully achieved. The first objective has been achieved through using boarding data from a fleet of bus vehicles that had an automated passenger counters (APC) penetration rate around 75%. The second objective has been achieved through many improvements in the boarding equations. Among these improvements is the separation of the current overall weekday peak into a morning peak and an afternoon peak. The re-estimated boarding equations perform well in terms of both how well the models fit the data and how observed and in-sample predictions compare. However, serious over-predictions can occur at a small number of stops, and these cases of over-prediction need to be dealt with individually as part of the validation process in each application of T-BEST.