PRB Utilization Forecast

[justdream]

Dears,

Have you did before LTE DL PRB Utilization forecast?
Any idea, which tool/ technique to use to predict how much PRB Utilization for each cell at end of 2019

[jbada]

I suggest you to forecast using the RRC, Active UE better then PRB

[justdream]

I suggest you to forecast using the RRC, Active UE better then PRB
but our final objective is determining cells with High PRB utilization need expansion
so how Active UE, User number can help..

[coach]

Hi,

The way I have tackled this kind if problems is by using some data science methods, either with R or Python.

I suggest you to check the book from this post: http://www.finetopix.com/showthread.php/50885-Mining-Over-Air-Wireless-Communication-Networks-Analytics, specifically section 3.1 (Network Performance Forecasting Strategy).

Basically, follow this method:

1. Create a model to correlate traffic (RRC Users, Payload DL, Payload UL) vs PRB usage. You can enrich this model by also adding metrics such as Coverage (RSRP distribution), TA distribution or other metric that shouldn't significantly change over time. To create this model, use a machine learning library from Python or R (I use python's scikit-learn).

2. Forecast traffic using some statistical method (easiest would be linear regression, but there are also algorithms to forecast data series, such as ARIMA. Here is an interesting article on this subject: https://towardsdatascience.com/an-end-to-end-project-on-time-series-analysis-and-forecasting-with-python-4835e6bf050b

3. With the forecasted traffic data, and assuming the coverage and TA remain the same, use the model created in (1) to predict the PRB Usage.

The procedure evidently requires some researching on your side. I suppose that some network or third party vendors may sell some software that can do this kind of analysis, but this method is fairly simple and free to implement!

Hope this information is useful!

Regards,

[electron]

Hi,

The way I have tackled this kind if problems is by using some data science methods, either with R or Python.

I suggest you to check the book from this post: http://www.finetopix.com/showthread.php/50885-Mining-Over-Air-Wireless-Communication-Networks-Analytics, specifically section 3.1 (Network Performance Forecasting Strategy).

Basically, follow this method:

1. Create a model to correlate traffic (RRC Users, Payload DL, Payload UL) vs PRB usage. You can enrich this model by also adding metrics such as Coverage (RSRP distribution), TA distribution or other metric that shouldn't significantly change over time. To create this model, use a machine learning library from Python or R (I use python's scikit-learn).

2. Forecast traffic using some statistical method (easiest would be linear regression, but there are also algorithms to forecast data series, such as ARIMA. Here is an interesting article on this subject: https://towardsdatascience.com/an-end-to-end-project-on-time-series-analysis-and-forecasting-with-python-4835e6bf050b

3. With the forecasted traffic data, and assuming the coverage and TA remain the same, use the model created in (1) to predict the PRB Usage.

The procedure evidently requires some researching on your side. I suppose that some network or third party vendors may sell some software that can do this kind of analysis, but this method is fairly simple and free to implement!

Hope this information is useful!

Regards,

Hi There,

Looks very interesting , Any document, study, macro, sample workspace for calculation to get the lesson learnt ?


B
R

[mounir34]

Dear
Please any guidelines can support for analysis, thanks