That’s the correct example.
Because I have seen all the garbage of Nokia, Ericsson etc applying ML and DL on Telecom KPIs.
Without even looking at the statistical distribution of the telecom data.
Anyone who has studied basic data science can tell that whatever Nokia Ericsson etc are selling in name of Data Science solutions is absolute garbage.
I have been looking into such solutions since 2018.
And till now, there’s literally no improvement.
Ok.
If i say drop call rate is 4%, what do you mean by this?
It means nothing… Because it’s perfectly possible, there were only 10 calls in the cell.
It’s incomplete and wrong number.
The right way is to say, drop call rate is 4%, for every 100 calls on the cell, irrespective of how many hours it took to make 100 calls.
In this way, you can compare any cell in the network.
Without the need to add qualifiers like urban rural areas, busy hour, non busy hour etc.
It’s all included if the KPI takes the number of drops for every say 100 calls.
Don’t you understand the concept of busy hour?
And why it was actually introduced?
That’s also wrong concept.
Because for a highway cell, busy hour may mean 10 calls in the hour.
While for a city cell, it may mean 2000 calls in the busy hour.
How do you compare drop call rate of BH for two cells.
You are just saying the same thing… 4% means 4 drops out of 100 calls.
Yes.
But telecom operators calculate the number of drops divided by number of calls per hour.
No, not like that.
This ‘per hour’ thing is wrong.
It’s not per hour.
Depends on the report that you check.
It’s always for some time domain.
Like one hour, or one day, or one week or one month.
Time is always in the denominator.
And this is where the mistake is.
No such per hour thing except busy hour.
Time is the 4th dimension.
Of course system always generate report for a certain time interval.
If you take 24 hrs all kinds of events are included.
And busy hour is to estimate the capacity of the system.
Anyhow I have said what needs to be said. You may not be able to understand it unless you study Data Science.
For Poisson distribution, the denominator must be frequency.
We need to work in frequency domain.
Like we cannot explain Prism phenomenon in time domain.
Can you explain the Prism phenomenon in the time domain?
How light is split into several colors?
If you can, keep reporting telecom KPIs in time domain.
Time and frequency are two different things.
Of course the same concept can’t be applied everywhere.
While things can be analysed in time and frequency domain, each has its own significance.
Poisson distribution cannot be analyzed in time domain.
The worst days in Telecom are not over yet.
So if you are looking for an overseas job, Below companies may be a good starting point to for your job search.
Qualcomm to cut more than 1,000 jobs – report
https://www.reuters.com/technology/cisco-lay-off-5-workforce-2024-02-14/
