Active User/Device Definition

[hakkiToran]

Problem description The current project documentation on the Region User Count project page defines "active users," but the definition is unclear. Specifically, it is not clear whether "active users" include only those actively using data or if it also considers users who are connected but idle. Please check this website : What is the definition of "active user"? Does an "active user" include users in an IDLE state as described [here] https://worldtechie.blogspot.com/2013/04/lte-ecmemm-and-rrc-states.html

Possible evolution Provide a detailed and clear definition of "active user" within the documentation. Specify whether it includes both users actively using data and those connected but idle.

Alternative solution Include examples or scenarios illustrating the "active user" status to enhance understanding.

Additional context

Clarifying this definition will help users accurately understand and utilize the API, ensuring better implementation and integration.

[chinaunicomyangfan]

@hakkiToran Hello,Welcome to join our discussion. We have discussed the definition of active users here and ultimately determined that the number counted by this API is the number of connected devices. As long as they are connected, they are counted regardless of whether they are idle. We have also decided to change the API name to Region Device Count. This has been submitted to the GSMA and apibacklog working groups

[hakkiToran]

Actually except from the definition, If you query the number of User for time=t for, what you will get ?

You will get the non-IDLE users. However the problem is most of time the users will be on IDLE for a specific time=t. Maybe only %20 of the users will be non-IDLE. Because if you leave your phone on table, maybe after 10-15 seconds, the phone will change to IDLE state. That's why querying active users for a specific time=t gives you may be maximum %20 percent of the users. Most of the time, the UEs will be on IDLE state, if not actively a person is using in hand. For example in earthquake scenario, most probably, the specific time=t based queries will miss to cover more than %80 of users.

https://worldtechie.blogspot.com/2013/04/lte-ecmemm-and-rrc-states.html

How can wesolve this problem? We need to query for a time interval instead of quering for a specific time. If time interval for example 15 minutes or 1 hour is used, the result will most probably show the more than %95 percent of user. Because UE and network making background signaling peridically like paging cycles, TAU times.., so in this time interval the phone must have made activities even in the pocket or on table . So the the time interval based query cover more users.

[gregory1g]

Let's change terminology to devices in this discussion. Ideally starting from the title of the issue.

[hakkiToran]

@gregory1g I changed the name as user/device. Changing the name for this issue does not eliminate the not getting the IDLE users .

How can wesolve this problem? We need to query for a time interval instead of quering for a specific time. If time interval for example 15 minutes or 1 hour is used, the result will most probably show the more than %95 percent of user/device. Because UE and network making background signaling peridically like paging cycles, TAU times.., so in this time interval the phone must have made activities even in the pocket or on table . So the the time interval based query cover more users/devices in the region, which will be more correct.

[chinaunicomyangfan]

@hakkiToran In my understanding, as long as the device is turned on and connected to the network, regardless of its status, it needs to be counted. For idle devices, regular location update signaling can be used to count. So I believe that the number of connected devices can be counted using time points or time windows

[hakkiToran]

@chinaunicomyangfan You are right, when a time interval based query made like for example (T-xx minutes ) where xx configurable can be much more useful. The value of XX should be for example 15 < xx < 300 minutes. Beca One important point to consider with time interval based (T-xx minutes ) query is to get the last cell. Because some users may be moving under different cells. So the API should return the last cell during the time interval.

[gregory1g]

@hakkiToran , this sounds like last known location in [T-xx; T] time frame. And sounds like a reasonable definition/approach to me.

[hakkiToran]

@chinaunicomyangfan and @philipxujin Will you implement this suggestions to the code ?

[chinaunicomyangfan]

@hakkiToran Sorry for the late response. We have been busy with the Shanghai MWC recently. We will carefully consider your proposal and add it to the code. It is expected to be completed next week