Predicting YouBike

Modeling demand for Taipei area bike share stations

Transcript

1. Predicting YouBike YouBike 預測 youbikeforecast.com 劉知岳 (Brandon) brandon@pureinformation.net
2. Why Forecast?
3. UNDERSTAND Observe past phenomena confirm or deny intuitions based on data quantify service quality
4. MAKE DECISIONS See what is likely to happen in the immediate future
5. What applications? http://pureinformation.net/transit-tracker/
7. Public YouBike API http://opendata.dot.taipei.gov.tw/opendata/ gwjs_cityhall.json
8. youbikeforecast.com
9. TOOLS LevelDB - data store Go, nginx, python, EC2 - server Leaflet, d3 - web page
10. The shape of the data tells a story
11. How might we classify YouBike stations based on usage patterns?
12. Pattern: Morning/Evening Commute （信義／杭州路⼝口）
13. Pattern: Night Owls （國⽗父紀念堂）
14. Pattern: School （公館（台⼤大前⾯面）
15. Choosing a metric What can we measure that affects users? How many minutes per day you can’t rent a bike (or what % of the day)
16. Pricing change March 2015（師⼤大）8% Unavailable
17. Pricing change April 2015（師⼤大）3% Unavailable
18. Improve the model? Weekend/Weekday （信義／杭州）
19. Improve the model? Weather （師⼤大） 下⼤大⾬雨
20. Adjustments
21. github.com/bdon/youbike-data