Hi Hussein, I remember we were discussing techniques for variable selection and dimension reduction in an earlier thread (). Do some of those techniques give you a better model than others? Different techniques have different pros and cons. Many times you have to try a lot of them to see which one is most helpful to your data and business problem. The course Advanced Predictive Modeling Using SAS Enterprise Miner is a good hands-on refresher of all the tools you have in your toolkit. I am considering taking that class again :smileysilly: Post some of your findings if you have a chance. Thanks! -Miguel ... View more

Hola Nico, Español está bien, pero no juzgues mi ortografía, de acuerdo? En este tip (How to build a credit scorecard) comentaba que la forma ideal de reintroducir variables importantes en tu Scorecard es ajustar manualmente el Weight of Evidence. Nota que antes de ajustarla manualmente el IV era de menos de 0.1 y quedaba excluída del Scorecard. Después de ajustar el WoE basado en conocimiento del negocio, hasta que quedó con IV mayor de 0.1. Otra alternativa es mover el Information Value cutoff en el Interactive Grouping node (lo encuentras en las propiedades de Score casi hasta abajo). No obstante, si aún quieres ver la matriz de correlación, ve este paper (página 7-9). Tienen un muy buen ejemplo que hace justo eso. Saludos! -Miguel --- summary in English There is a best practice to reintroduce a variable that does not meet the cutoff for information value or gini. Override the WoE as described in this tip: How to build a credit scorecard. If you still want to produce a correlation matrix, this paper describes how to do it with a SAS Code Node. Look at pages 7-9. ... View more

HI JBerry, Thanks for the detailed question!!! Nice screenshots and awesome profile pic BTW! I borrowed this idea from Wendy Czika. Do this: 1. Add a Score node after your Survivial node and run it. Open the results and copy the Optimized code. 2. Connect a SAS Code node after your Data set. Open the editor and add this: -libname statement to create a library (remember that valid library names are 8 characters or less) -data statement to output your results -set statement for your data set -the optimized score code you copied from the Score node -run statement It would look something like this: libname results "D:\EM\EM_Projects\EM13.2\miguel"; data results.scored36m; set &EM_IMPORT_DATA; /* your optimized core code goes here */ run; 4. Scroll all the way down to the last part of the optimized score code you pasted. Right before the end you will see the part where the macro EM_SURVEVENT calcualtes the survival probability. Add the code highlighted in yellow. It creates the variable IntervalsInFuture and outputs the calculations for all periods after _t0_. /***** omitted lines of code ******/   if _T_=t0_fcst then EM_SURVEVENT=(EM_SURVIVAL-EM_SURVFCST)/0.00001; end; /*just to be able to easier see how many months after the censor date we are looking at */ IntervalsInFuture = _t_ - _t0_; /* output each period */ if _t_ >= _t0_ then output; _t_+1; end; _T_ = _T0_; ; end; 5. Open your results data set to confirm this worked as intended. I just tried this with my go-to example and it worked great. I hope this helps! Thanks, https://www.linkedin.com/in/migmaldonado ... View more

OK, it took some googling, but I got this working. You have a couple options. 1) you pass the train, validation, and test sets using the macro variables so that a Model Comparison node can pick up the partition and calculate the stats. 2) take advantage of the specific proc syntax. I think this is what you were trying to do: 1. Add a data set. In my example I used German Credit from F1->Generate Sample Data Sources 2. Add a Partition node 3. Add a SAS code node with the code below. Change the bold for your own target (response) and inputs (effects). data mydata; set &EM_IMPORT_DATA(in=a) &EM_IMPORT_VALIDATE(in=b) &EM_IMPORT_TEST(in=c); if a then _partition="_Train"; else if b then _partition="_Valid"; else if c then _partition="_Test"; run; proc glmselect DATA=mydata; effect MyPoly = polynomial(duration checking savings/degree=4); model amount = MyPoly; partition rolevar=_partition(TEST='_Test' TRAIN='_Train' VALIDATE='_Valid'); run; 4. Run. The output results will give you the ASE of training/validation/testing. This model isn't fabulous for this data set but hopefully this approach will give you good results on yours! Good luck! -m Good reference: SAS/STAT(R) User's Guide, proc glm select - partition statement PS If you try the other approach I described, you can easily use a Model Comparison node to compare with HPGLM or any of the model nodes in Enterprise Miner. Very recommended to give this a try!!!!!!!!!! ... View more