Number of users change over time

[hakkiToran]

Problem description
The number of users change over time can be a valuable output. So the API should also give the number of users in in an area with time dimensions for the below scenarios.

what is the user count between 10:00 and 14:30 on 13th of May
Time window can be set:
 HOURLY: (user count between at 11:00 on 13th of may )

 Period of the Day : morning, midday, afternoon, evening, night ((user count between morning 13th of may )
For example in google MAPs, I can find the weekly and hourly change of visitors for a certains locations.

Possible evolution

API can query with time interval parameter and output the result with hourly distribution.

The API output can give the date or day of week and hour information in the output.

Alternative solution

Additional context

For example when I check a place in google MAPs, I can see the hourly visitors count in a day and also visitors distribution with day of week distribution like monday, tuesday...

[chinaunicomyangfan]

@hakkiToran I think this is a great proposal. If we want to meet the scenario you mentioned, we may need to add a new endpoint to query the number of devices in a certain area during a certain time period. At the same time, this endpoint can also meet the needs of User Story 2- outdoor anchors can infer future data based on historical data and choose suitable live streaming locations.
I think users can input a time point or time window. If a time point is passed in, the number of devices at that time point will be returned. If a time window is passed in, it is reasonable to output the results in an hourly distribution.
Additional explanation: The time window should need to be limited. Based on the actual situation of our interface, data within 7 days can be accessed, so it should be reasonable to limit the time within 7 days

[hakkiToran]

@chinaunicomyangfan I am agree with you about adding an endpoint for the query. In #21 I also mentioned that we need to add an time interval for our queries. Because IDLE user ratio be very limited for a specific time=t, because if there is no activity in phone, the phone will go to IDLE state may be 10-15 seconds later. Instead Time interval can be 10 minutes, 15 minutes or 1 hour.

I am agree with you that time window can be limitted. With 7 days we can see the change of user count or intensity on a region for different days of week. Also with the hourly distributions we will see the hourly user count for different hours of day.

[gregory1g]

Same comment as for #21 - lets use terms we agree about.

Regarding the topic: there are two possible approaches here:

1. API defines a time window which is considered. For example if a request is done for time T. The API returns number of devices in the area detected in the interval [T-15min; T].
2. API consumer can specify time interval (and API must specify precision to be supported).

The second option must be double checked from the privacy point of view - if one can request overlapping intervals to bypass privacy threshold. I tried different attacks scenarios in my mind, but did not found any working one. But I can miss something here.

[hakkiToran]

@gregory1g I am agree on using time interval based query. We can use use a time interval based query like [ T - xxmin ] where x is configurable between a range of 15-120, or 15 to 240 minutes.

When we use the a time interval based query, in order to avoid the multiple cells to be reported, the last cell should be considered. (The device/subscriber may have multiple cells. In this case last cell during the [ T -xx minutes] should be used.)

[hakkiToran]

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

[chinaunicomyangfan]

@hakkiToran The function of querying through the time window has been merged here. Can we close this issue？

[hakkiToran]

@chinaunicomyangfan Thank you for the update. Yes we can close this issue.