Using GPS to Learn Family Routines: A Summary
Download as PPTX, PDF1 like366 views
The document discusses using GPS and machine learning to improve family routines by making them computable, addressing coordination issues that lead to anxiety and missed activities. Key contributions include modeling routines for better planning, visualizing daily schedules, and accurately predicting drop-off and pick-up times. The system demonstrates high precision and recall rates in predicting transportation needs, ultimately enhancing family organization and reducing stress.
Using GPS to Learn Family Routines: A Summary
- 1. Using GPS to Learn Family Routines: a SummaryScott Davidoff
- 2. Everyday routines are unavailable to most sensing systems
- 3. Smart phones can make routines computable
- 4. Why does that matter?
- 5. Use routine models to lower anxiety when juggling kids’ rides
- 6. Identify mechanisms that result in coordination problems for familiesCONTRIBUTION 1
- 7. Demonstrate a new activity lies in the scope of sensing + machine learningCONTRIBUTION 2
- 8. Visualize the day’s plan using learned activity, place, pick-up time and driverCONTRIBUTION 3
- 10. Routines are not documented on calendars or elsewhere1
- 11. Families do not recall one another’s routines perfectly2
- 12. Families make plans that depend on incorrect information3
- 13. Why is this a problem?
- 14. Sometimes kids get left at activities. Anxiety for everybody.
- 15. Contribution 2
- 16. Sense pick-ups and drop-offs1
- 17. Sensing Drop-Offst4t3t2t1PLACE 1PLACE 2ChildParent
- 18. Sensing Pick-Upst4t3t2t1PLACE 1PLACE 2ChildParent
- 19. Over 90% precision and over 90% recall without supervision
- 20. Predict drivers2
- 21. Use a decision tree:distribution over past driversreal-time locationobservables
- 22. Unsupervised, online learning using 1 week of training data is 72% accurate
- 23. Unsupervised, online learning using 4 week of training data is 88% accurate
- 26. Time to Pick-UpA’ Value0.6590.8010.825tideal – 30tideal – 10tideal
- 27. Contribution 3
- 28. The Family Time-Flow ® shows the plan for the day using routine models