Skip to content

R package for scorecard development

3 messages · Nikhil Abhyankar, Suzen, Mehmet, Michael Friendly

#
Hello all,

Is there any R package that can develop a scorecard model for a binary
target variable?

More details:
I want to create a scorecard based on the raw data I have.

I have a binary target variable and a few numeric and character input
variables.

I want to bin the variables and assign a score to each of the bins.

Each subject will be scored based on the bin it falls in for each variable.

All such scores from each of the variables will be added up to get the
final score.

There will be a cutoff score to decide which of the two classes of response
the subject falls into.

I fount and tested the smbinning package. However, it only gives the bins
for a single variable at a time.

How can I get a full scorecard model?

Thanks
Nikhil
#
I suggest you to have a look at this R document:
https://cran.r-project.org/doc/contrib/Sharma-CreditScoring.pdf
On 28 June 2017 at 13:26, Nikhil Abhyankar <nikhil.abhyankar at gmail.com> wrote:
#
Hi

I'm sure there are valid reasons for wanting to use a scorecard model,
but a more straightforward approach would just be a logistic regression
or logistic discriminant analysis.

Compared to that, a scorecard model can be considered to be throwing 
away information by binning the predictors.  It is similar to what scale
developers often do by simply summing up item (0/1) scores on the 
assumption that they all should be equally weighted, rather than using 
something like factor or component weights.

Just a thought.  If you do come up with a scorecard model, it would at
least be useful to compare it with a logistic model.

-Michael
On 6/28/17 1:26 PM, Nikhil Abhyankar wrote: