Discussion on API algorithm - Initial proposal

[jgarciahospital]

Initial proposal of API algorithm for

• Data management
• Extrapolation
• Prediction

Draft Proposal

Open for discussion.

[timrawling2]

Output 4: Is this output a square grid, or could it be describer by e.g. h3 hexagons? Where is the size of the grid requested or defined?

Step 5: In addition to this step prioducing the number of UEs per grid, should this also attempt to classify UEs as being static or moving? Depending on speed, this could allow an estimate of the number of people who are stationary (likely indoor), walking or in vehicles.

Step 6: Should each grid also have a 'variance' value associated with it? The population density forecast for next Tuesday should be similar to last Tuesday, and the Tuesdays before that. However, there may be grids that display a wide variance for that time slot. Also, should the output for a future period represent an average of expected UEs based on historical information, or a maximum?

Step 7: Agree that the ultimate output should be a total population per grid, and that this would initially be based on market share. However, the results of step 7 could be improved by querying multiple MNOs and adding together the responses. Should the algorithm allow for output following step 6, and multiple inputs to step 7 to allow this?

Output 7: Should this be a total population per grid, or would returning a population density figure, or a SAIL category be better?

On the point that from "From step 2 onwards, this is aggregated data that does not contain personal information (this is anonymous data).", this is true. However, there are still privacy concerns - if a user lives in an isolated building, the API could be inadvertently releasing personal information on whether that user is at home or not.

[bigludo7]

Hi @jgarciahospital Got a small comment....probably worth to mention for the point 2 a A cell is associated with each user in each time interval that it could be challenged locally depending on local regulation.

Pseudonymized input data as described may not be directly legal in all countries. In France, to ensure legality, the maximum duration is 15 minutes for silent mobile - which is fine with the proposal but not sure this is the case everywhere.

[VGF-TEF]

Thank you for bringing this up. The 15-minute interval is a guideline and can be adjusted to suit the specific geography regulations. The idea is to establish a time interval that aligns with the MNO’s capabilities and the legal requirements of its respective country.