Information regarding passenger volume is incredibly important for train operators. It is used for scheduling timetables, determining ticket prices, distributing revenue between service operators, planning rolling stock, and in reports.
Today, Automated Passenger Counters (APC) are available on some of vehicles. However, trains often run without an APC. This can be for number of reasons, such as lack of installation, equipment being out of order, or data quality being unsatisfactory. In such circumstances, a statistical model that estimates the number of passengers boarding or exiting the train is needed.
Automated passenger predictions
We have developed an analytical system for passenger volume estimates in close collaboration with Vy. The system is based on data collected from active APCs. The system is autonomous and produces daily predictions for passenger numbers on all arrivals and departures at each station for thirty days.
As an APC is not always installed or measured, the system also predicts missing counts. The model uses statistical tools to incorporate trends and other type of covariates to account for systematic variation in the data. It also takes into account seasonal variation, in addition to yearly, weekly and daily patterns. Patterns are in constant flux, and the system draws on new data continuously in order to reestimate the model and provide accurate, up-to-date predictions.
Name: Prediction of Vy’s passenger traffic based on APC data
Period: 2013 – ongoing