Roll Rate model, as I observe, has long and high popularity among business
managers, especially those with MBA background, finance guys, banking
executives.... Some participants in this SAS group may not agree that your
original question has much relevance to what this group is mainly about.
Roll Rate model often centers around spreadsheet and is prepared by
financial reporting artists.
However, what you are after may be relevant. Non-linear models, cluster
analysis, time series analysis, to name a few, are all at the core of
SAS/statistics, ETS or OR. Classical regression analysis, logit and others
enjoy popularity as well. Those models potentially can keep forecasting
error margins around 3-5%. Given, say, your current balance every month can
be billions, to keep your cash reserve down, 3-5% may still be too costly.
To further improve, other models such as neural networks, fuzzy logic,
machines, multilayer-prediction, sometime decision trees are relatively more
precise. Those models demand very sophisticated data preprocessing. SAS, I
believe, is the best tool to do that.
I am still struggling with the idea of licensing SAS Enterprise Miner. I am
not able to tell you really what EM can do for you in forecasting. I have
been using SAS Base to preprocess data for KnowledgeStudio and have been
happy with the results in forecasting loan loss, charge-offs, attritions,
roll-overs, mail bag, solicitation, cross-sell and so on.
Methodology wise, one approach is to try to score risk on each card account
and then resolve aggregation over the accounts. Aggregating over a tradition
accounting ledge can be pitfall. This approach, on the micro level, involves
robust scoring models. Another approach is relatively quantitative. If you
have series of many months accumulated, time-series is sometime very good.
If you ling towards actions/decisions, you may consider decision trees.
Generally, stochastic models seem to catch on recently. There is a magazine
deseased by the end of 1999 called NeuroVest, out of Virgina, dealing with
these type of models. Their focus was on stock market prices and portfolio
management. They are concise but very technical. Since your question is
Financial, it is highly relevant. So far I am a semi-newbie in SAS, so I
don't have a lot to "brag" about in terms of how to use SAS to apply those
exciting algorithms. Hopefully somebody else can relate more to SAS.
Happy Chrismas.
Paula D
"Anna Henson" <[EMAIL PROTECTED]> wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> Happy Holidays Everyone,
>
> I am interested to know if there is any good forecasting models for profit
> and chargeoffs for credit card industry besides the roll rate model. Any
> help is greatly appreciated.
>
>
> Sincerely,
>
> Anna Henson
>
>
>
>
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
