The traditional method of obtaining a statistical portrait of transit users’ commuting habits was to conduct one-time surveys, which provided only part of the picture. Today, when the microchips in transit passes are scanned, this generates a wealth of data. Researchers can sift through those data and use them to improve transit network service. This exploratory approach is an integral part of the research project we are conducting in partnership with SNCF-Keolis.
“When you use a chip card to get on the métro or the bus, it creates a set of data that says, ‘this card was scanned at such-and-such a place and time,’” explains Maguelonne Chandesris. Over time and with larger populations of riders, these data accumulate. To extract information useful to decision-making, SNCF-Keolis partnered with the Interuniversity Research Centre on Enterprise Networks, Logistics and Transportation (known by its French acronym, CIRRELT), one of our academic members. Using data from the Société de transport de l’Outaouais, among others, they are developing algorithms to gain better insight into public transit network management.
As a first step, the team sought to identify families of rider behaviours. Consider, for example, and a priori, daily commuters who take a bus from Monday to Friday, and all other possible transit rider routines. Keep in mind, however, that an algorithm has no a priori considerations built in. The researchers therefore developed an analytical method that groups together, with no a priori considerations, chip-cards that follow a similar usage pattern. Applying the method month after month, they succeeded in determining and tracking the relative sizes of these families of behaviours, and were able to observe, for example, whether the group of daily riders that was dominant in January was still dominant in July, when most workers go on vacation.
With the first method, we note that the group of riders is smaller in summer, and with the second, we see that this group changes because fewer people are working on Fridays.
In July, however, some workers who do not go on vacation adopt a reduced schedule of four days per week. In that case, the composition of the group of daily riders changes compared to what it was in January. This raises a broader question: Do the families of behaviours retain their characteristics year-round? To answer it, the researchers therefore developed a second method to track changes in user behaviours.
We’re very happy to be working with the team at the CIRRELT, a state-of-the-art, pioneering group that for years has been working on chip-card data.
Business Developer with Kisio, a business unit of Keolis, groupe SNCF
Chip-card data still have a lot to tell us, and we will be continuing our research to study how behaviours change to reflect events such as festivals and international sports competitions, as well as to categorize métro stations by traffic volume. Over the longer term, SCNF-Keolis plans to apply these methods to the SNCF’s Transilien suburban rail network in the Île-de-France region, to better understand user behaviours and, ultimately, improve service.