To process the mobile phone data, KOTI applied the following steps:
1. Raw data: Anonymised mobile phone records giving user id, mobile base station coordinates, time at origin, time at destination, duration of stay, gender and age group.
2. Estimate duration of stay at each mobile phone base station: Point data are first ordered in time sequence. The initial and final times of communication with the mobile device are recorded, which then allows to estimate the duration for which the mobile device is identified at the base station.
3. Temporal and spatial data correction: Since mobile phone data only shows which cell tower a device is connected to, and only when the device is used, temporal and spatial corrections are applied.
4. Temporal: Estimate the actual position of a user based on the frequency of record at each base station for each time slice.
5. Spatial: Estimate the actual position of a user based on the stay points and duration of stay, for points within walking distance.
6. Identify area of longest stay:
7. For each user, estimate the list of locations and corresponding durations of stay over a week, and identify the location of longest cumulative stay.
8. For each day of the week, estimate the list of locations and corresponding duration of stay during a day and identify the location of longest stay.
9. Identify area of second longest stay:
10. Reiterate the steps above excluding longest stay location.
11. Estimate the type of the main stay area:
12. Label stay points using the criteria in Table A C.1.