Developing and Validating Models in Telecommunication Sector with SAS?

[ertr]

Hello everyone,

I would like to learn about which reports do the telecommunication companies use for validate their models? As you know, at the Risk side,they look for Gini, Binomial, PSI, Herfindahl and etc.? And they develop Probabilty of Default models. But what they use in telecommunication?

On the other hand, I wonder that what kind of models do they develop? As far as I know, they develop Churn models? Anything else?

How can I develop or create these reports and models in SAS? Is there a best practice for this in SAS? Or any examples?

I am open to listen your suggestions or directions.

Thank you.

[Reeza]

Have you looked at any of the SAS EM book, primarily the one titled practical applications?

[ertr]

Do you mean Applied Analytics Using SAS Enterprise Miner? In which part does it mention about it?

Can you be more specific?

Thanks

[ertr]

I found this and this one is useful and my purpose is to examine these kind of papers and methods deeply. Any suggestions?

http://www2.sas.com/proceedings/sugi27/p114-27.pdf

Thanks

[Reeza]

https://www.amazon.com/Predictive-Modeling-SAS-Enterprise-Miner/dp/1629602647

Unfortunately, I cannot be more specific, I gave away my version to a colleague. 

[Capt_VA_SAS]

Ertr,

Telecommunications is alot like most other industries in that we developed all sorts of models for various use cases across the industry.  I myself was in the Credit and Collections group for a major carrier and we built logistic regressions, decision trees, and more for Probability of Involuntary Churn and linear regressions for Amount Past Due.  We also modeled for which collections method would work best.

The Marketing department would build Voluntary Churn models all the time as well as Cross-sell/Up-sell models.  They would also build recommendation engines for On-Demand video content as well as "Next Best Offer" models.

The Network engineers would use the Erlang and Gaussian distributions to model network traffic, but didn't really build any predictive models.

Finally, Operations would forecast network traffic, store foot traffic, customer churn, and sales monthly.