Predicting and labancing bikeshare systems by Raphaël Cherrier

Predicting and labancing bikeshare systems by Raphaël Cherrier

Presentation at DataGeeks @AXA avec Sean Owen January 21 2015· 7:00 PM
Raphaël Cherrier, PhD
QUCIT - Founder & CEO
www.qucit.com

B4c5e3be1957239def212083b3f50916?s=128

Datageeks Paris

January 21, 2015
Tweet

Transcript

  1. Predicting and balancing bikeshare systems @raphaelcherrier

  2. Disclaimer: Short Bio PhD in Theoretical Physics (Spin Glasses, Disordered

    Systems and Neural Nets) Associate Prof. in Fluid Mechanics Algorithmic Trading Founder @qucit (may 2014)
  3. Clément Data Science & Mobile Dev Msc in Physics (ENS

    Lyon) Yassine Data Science, Dataviz & Mobile Msc in Applied Mathematics (Centrale Paris & Cambridge) Raphaël Founder & CEO Statistical Modelling PhD in Physics (ENS Lyon) Michaël Partner & CMO Business Development (Centrale Paris & MBA Insead) Rémi Deep Learning & Cloud PhD Student in CS (Univ. Bordeaux) Nicolas Data Science & Cloud PhD in Mathematics (ENS Ulm) Caixia Dataviz Msc in Computer Science (Univ. Bordeaux) Guillaume Cloud & Graphs Msc in Computer Science (Univ. Bordeaux)
  4. BIKE SHARE SYSTEMS PROBLEMS CURRENT SOLUTIONS

  5. 700+ bikeshare systems www.bikesharingworld.com Started in France (2005, Lyon) 3b$,

    +50% per year 1M+ bikes 830+ cities THEY ALL SHARE THE SAME PROBLEMS
  6. None
  7. daily_trips(system_size) ~ N1.46

  8. None
  9. None
  10. CURRENT SOLUTIONS

  11. Real-time Availability

  12. Balancing with Trucks POLLUTES & emits CO2 NOT Efficient EXPENSIVE

  13. PREDICTING BIKE AVAILABILITY

  14. Raw Bikes Data

  15. Riders & Trucks

  16. Riders & Trucks Balancing !

  17. Riders & Trucks Balancing ! ‘Normal’ Trips

  18. Bike share stations are like mikes that listen to urban

    mobility behaviors, measuring the ‘pulse’ of the city ‘UNIVERSAL’ MOBILITY BEHAVIORS
  19. BEHAVIOR 1

  20. BEHAVIOR 1

  21. BEHAVIOR 1 = COMMUTING

  22. Bordeaux Toulouse Barcelona New York

  23. None
  24. None
  25. None
  26. BEHAVIOR 2

  27. None
  28. BEHAVIOR 2 = GOING OUT

  29. None
  30. None
  31. None
  32. …are affected differently by contextual changes ‘UNIVERSAL’ MOBILITY BEHAVIORS

  33. Predictions up to 12h ahead

  34. AI for bike balancing & failure detection

  35. Optimal network planning

  36. We Are Hiring! @raphaelcherrier