Experiment design
The correct answers (ground truth) come from a planned field survey. Participants visited places with set instructions and wrote down where they were.
- 11 participants, each doing 2 sessions.
- 5 groups of stops (A to E), each testing one theme. Each session has 12 planned stops and 4 free stops chosen by the participant: Group A: vertical position (floor level), Group B: sky view obstruction (how much sky is visible), Group C: indoor position (position inside a building), Group D: movement type (how the person moves), Group E: ambiguous edge cases (places that are hard to classify).
- 3 phones per session, to collect more GPS data. The three recordings of one stop are not independent of each other. This is why testing uses Grouped cross-validation.
- The start and end of each stay come from the Self-report form, not from an algorithm.
Result: Ground truth dataset. Part of Ground truth data.