What query logs tell us about travel flows

We studied the query logs of the iRail route planning API to see whether it can serve as a way to study travel flows in Belgium. We want to know, because iRail wants to launch a dataset than tells you how busy your train is. In Belgium, this can help you tell on which trains you will still be able to get some work done, or read a book, instead of having to stand up.

Crowd-funding campaign: https://spitsgids.be/

Interactive origin-destination visualization: http://rxd.architectuur.kuleuven.be/irail/weekdays.html

Full paper: http://usewod.org/files/workshops/2016/papers/paper.pdf

iRail query logs:
- Historic: http://gtfs.irail.be/nmbs/querylogs
- Real-time: http://api.irail.be/logs

Transcript

1. What query logs tell us about travel flows Pieter Colpaert, Alvin Chua, Ruben Verborgh, Erik Mannens, Rik Van de Walle and Andrew Vande Moere
2. A route planning API http://api.{myapp}/?from={A}&to={B} &departuretime=2015-09-09T14:45Z &wheelchairaccessible=true &transit_modes=plane,railway,bus,car ...
3. Can we use the query logs of a public transit route planning API as an indication of travel demand?
4. Why would we want to know about travel demand?
5. Need for evidence based mobility policy Is the offer correctly adapted to the demand? * Source: planned train schedules GTFS at http://gtfs.irail.be ** Not taking into account the train composition: no data available
6. Or predictions may tell you which train to avoid
7. Or we are just supposed to know
8. Why query logs? Contain origin, destination and desired time of departure → Predict near future, not only current state Belgian railways don’t have check-in/check-out system Many tickets not digital (Go/Rail-Pass, student pass, ...) Query logs are cheap in comparison to mobility studies
9. The iRail project “A hackerspace for a better transport experience in Belgium” iRail hosts: • data dumps http://gtfs.irail.be • a route planning API for anyone to prototype with http://api.irail.be
10. The iRail route planning query logs Nowadays: ~400k route planning queries/month
11. ~100% increase in April Official app discontinued in favour of other official app → the iRail alternatives gained traction
12. Mostly done by these 2 apps
13. And more smaller ones Plenty of different apps and code snippets for different platforms… Github search:
14. What’s the max size of our sample? In best case: 1 query == 1 intention ~4*10^5 travel intentions captured /month Estimated n° of passengers per month in 2015: ~2*10^7 (source: statistics Belgium) ⇒ max 2% of travelers plan using iRail
15. Query logs available as open data (CC0) https://github.com/pietercolpaert/irail-route-planning-query-logs http://api.irail.be/logs
16. Time based experiments 2 examples
17. (1) Students affect query logs distributions, and actual travel demand
18. (2) People look up their train earlier on days we also expect more people to go home earlier
19. Origin Destination visualizations http://rxd.architectuur.kuleuven.be/irail/weekdays.html
21. Conclusion In the iRail query logs, we find patterns that correspond to real demand The origin destination visualization shows Belgium as a star-shaped network with Brussels at the center
22. Inconclusion Cannot tell how good of a representation this data is. Treat this data with caution. Offer deficit vs. the demand? Cannot tell as we don’t know train composition, and maybe people are looking up their train earlier?
23. https://spitsgids.be Let’s crowdfund predicting how busy your train is Now launched