IoT Analytics Predicts Bike Sharing Demand
•
1 like•1,120 views
This presentation shows how to apply advanced analytics to forecast the need for timely restocking of bike stations in Washington DC, by taking into account past history, weather, calendar, traffic conditions, and additional external information. A model is created to predict the restocking need of each station by the hour and based on the minimum bare number of input parameters. The minimum bare number of parameters, necessary to guarantee an adequate prediction accuracy turned out to be: bike station, current bike ratio at the station (# bikes/# slots), time of the day, and weekend vs. business day.
Recommended
Recommended
More Related Content
What's hot
What's hot (20)
Viewers also liked
Viewers also liked (20)
Similar to IoT Analytics Predicts Bike Sharing Demand
Similar to IoT Analytics Predicts Bike Sharing Demand (20)
More from KNIMESlides
More from KNIMESlides (7)
Recently uploaded
Recently uploaded (20)
IoT Analytics Predicts Bike Sharing Demand
- 1. Copyright © 2015 KNIME.com AG Advanced Analytics for the Internet of Things Rosaria Silipo, Phil Winters, Aaron Hart KNIME.com AG Rosaria.silipo@knime.com
- 2. Copyright © 2015 KNIME.com AG IoT 2 Household Energy Wearables Health City Industry
- 3. Copyright © 2015 KNIME.com AG TheChallenges • Handling very large amounts of data created over time • Forcing sensor-equipped objects (house or city) to learn, and therefore to become smarter
- 4. Copyright © 2015 KNIME.com AG IoT Illustration by CRISTINA BYVIK Use Public Data Please…. Energy Consumption Prediction Energy Profiling Restocking Strategies Geo-localization Traffic Predictions Anomaly Detection
- 5. Copyright © 2015 KNIME.com AG
- 6. Copyright © 2015 KNIME.com AG This Use Case Capital Bikeshare in Washington DC
- 7. Copyright © 2015 KNIME.com AG
- 8. Copyright © 2015 KNIME.com AG Sensors!
- 9. Copyright © 2015 KNIME.com AG The Business Challenge:
- 10. Copyright © 2015 KNIME.com AG Even MORE of a Business Challenge • Any Station without bikes for 1 hour: $XXXX Per Violation • Any Station with no free slots for 1 hour: $XXXX Per Violation
- 11. Copyright © 2015 KNIME.com AGhttp://bikeportland.org/2013/03/10/behind-the-scenes-of-capital-bikeshare-84006 Capital Bikeshare Response Over 3 years 307 Stations 2963 Bikes 19.4% Casual Bikers 5.9m Bike Moves
- 12. Copyright © 2015 KNIME.com AG Advanced Analytics for the Internet of Things Pre-processing Data Visualization Predictive Analytics
- 13. Copyright © 2015 KNIME.com AG
- 14. Copyright © 2015 KNIME.com AG The KNIME Platform: Open for Innovation Powerful: Legacy Future Tools Collaborative: Scientists Analysts Integrative: Legacy Future Data Transparent: Existing Future Expertise Agile: Internal External Wisdom 14
- 15. Copyright © 2015 KNIME.com AG The KNIME Analytics Platform 15
- 16. Copyright © 2015 KNIME.com AG From Access to Visualization and Deployment
- 17. Copyright © 2015 KNIME.com AG
- 18. Copyright © 2015 KNIME.com AG Reading all Sensor Data
- 19. Copyright © 2015 KNIME.com AG Topology / Elevations Weather Holiday Schedules Commuters Enrich
- 20. Copyright © 2015 KNIME.com AG Elevation from Google API Connecting to Google API and other REST services available on the Web
- 21. Copyright © 2015 KNIME.com AG Expand/Transform
- 22. Copyright © 2015 KNIME.com AG
- 23. Copyright © 2015 KNIME.com AG Station and Bike Facts Over 3 years 307 Stations 2963 Bikes 19.4% Casual Bikers 5.9m Bike Moves
- 24. Copyright © 2015 KNIME.com AG Stations: deficits and surpluses
- 25. Copyright © 2015 KNIME.com AG Top 250 Routes
- 26. Copyright © 2015 KNIME.com AG Total Bikers Number per Hour • Registered (blue) vs. Casual (red) Bikers
- 27. Copyright © 2015 KNIME.com AG
- 28. Copyright © 2015 KNIME.com AG The Goal • Restocking alert signal • 1 hour warning! Lag(flag-1)
- 29. Copyright © 2015 KNIME.com AG Input Features • Weather related features • Number of registered and casual people showing up • Station infos (name and max. number of docks) • Calendar infos (working day, holiday, date) • Past infos • Number of bikes added and removed at each hour • Adjusted cumulative sum = number of bikes available at the station at a given hour • Bike ratio = adjusted cumulative sum/total docks available
- 30. Copyright © 2015 KNIME.com AG Restocking Alert System Train Apply Evaluate Partition 78% accuracy
- 31. Copyright © 2015 KNIME.com AG Feature Elimination Loop Two options: 1. Use all the input features (no thinking required, just a powerful machine) 2. Select the most useful input features via the “Feature Elimination” loop At each step, one input feature is removed - i.e. the input feature whose removal produces the smallest error increase.
- 32. Copyright © 2015 KNIME.com AG Input Attribute Impact
- 33. Copyright © 2015 KNIME.com AG Lean Restocking Alert System The input feature subset with the smallest error (81% accuracy): • Hour of the day • Working day (Y/N) • Current Bike Ratio • Terminal (station code) Past of bike ratio and weather infos do not seem to be relevant! Is this because most bikers are registered members?
- 34. Copyright © 2015 KNIME.com AG Lessons Learned • Enrich / Expand – KNIME Transformations – REST calls to external sources • Explore – Visualization with Open Street Map integration – Network Analysis (Graph Visualization) • Prediction – (Lean) Restocking Alert System – Weather influence does not seem to be important!
- 35. Copyright © 2015 KNIME.com AG Resources • KNIME (www.knime.org) • BLOG for news, tips and tricks(www.knime.org/blog) • FORUM for questions and answers (tech.knime.org/forum) • EXAMPLE SERVER for example workflows • LEARNING HUB (www.knime.org/learning-hub) • KNIME TV channel on • KNIME on @KNIME • KNIME on https://www.facebook.com/KNIMEanalytics 36
- 36. Copyright © 2015 KNIME.com AG Where can I find all this? Slides: http://www.knime.org/files/bicycle_final.pdf White paper, Workflows, and Data is available on the KNIME web site: http://www.knime.com/white-papers