topic Enterprise Miner: Collect score code from different branches into one score file in SAS Data Science

Enterprise Miner: Collect score code from different branches into one score file

ZoltanVincze — Thu, 23 Feb 2017 17:36:55 GMT

Hi all,

I have built a stratified logistic regression model. My stratas differs both in available data and sample size that's why I did not use the "group processing facility". Instead I built a process flow consisting of separate (parallel) branches:

Input data (shared among branches) -> Filter -> Interactive Binning -> Logistic Regression -> Score -> Score Code Export

for each of my stratas. My problem is that this process flow generates different score files for the different stratas. I would like to generate only one score file containing the score code from all branches appended one below the other. I tried to merge the different branches both at the Score node and at the Score Code Export node but it did not work. Only the score code of one branch got exported.

My question is: how can I merge the branches into one score code file / is there a better way for building the model?

Thanks in advance!

Re: Enterprise Miner: Collect score code from different branches into one score file

MelodieRush — Thu, 23 Feb 2017 22:18:09 GMT

I'm not sure I totally understand what you are doing, but if I connect all my models to a Merge node (under Sample tab) and then put on a score node, the score code is stacked for all of models into one file.

Is this what you are looking to do?

Re: Enterprise Miner: Collect score code from different branches into one score file

ZoltanVincze — Fri, 24 Feb 2017 10:10:11 GMT

I would like to build something like the second "stratified group processing facility" branch. As far as I know the property setting for the Interactive Binning node can not be changed between different runs of the group processing facility. But this is exactly what I would like to do. For example for a run processing less than 10k data rows I would like to set the "Number of Groups" property to 4 and for a run processing more than 150k data rows to 10. That is why I built the first branch using the Split node. In this case I have total control over the property settings. My problem is that I can not merge the resulting two models into one score file. As seen in the picture one of the 3 solutions runs on an error and the others export only one model.