09-19-2015 07:43 AM
If you are keen you could write your own IML randon forest code, but I think I saw a Rapid Predictive Modeller module in the latest SAS-U edition. The advanced setting would honestly (i.e. SEMMA training/validation comparison to ensure generalisability) compare lots of predictive regression models quickly and select the best. If you're not fixed on random forest the RPM module should generate useful predicitve models.
09-28-2015 03:47 PM
The Rapid Predictive Modeler is not included in SAS University Edition.
You may want to see the following thread to explore how you could implement some machine learning techniques using the procedures from SAS/STAT which is included in SAS University Edition:
03-08-2016 06:37 AM
Thanks for the correction, Sharad. I confused SODA and SAS-U Studio capabilities there for a second. I've used the RPM task in SODA Studio quite successfully on a number of direct marketing datasets for learning-by-doing in my marketing analysis and data analytics classes. It is a pretty powerful task. Maybe that could be the next new task on the wish list for the next SAS-U version?
03-14-2016 06:03 AM
..and just to further clarify - although there is no rapid predictive modeller task in SAS-UE (as there is in SODA) - there is a predictive modeller task. It does not fit all the m/c learning + stats models that the SODA RPM task does, but it still does a pretty good training/validation/test data split + honest asessement of linear models with button controls for main effects, nested, crossed, n-way interaction and full factorial predictor design settings, automated model selection using honest assessment and the ability to score new data sets from the automatically selected best predictive model. It uses GLMSELECT and REG. You'd have to try that first, before looking into more time-consuming approaches in SAS-U, right? An irony is that the predictive modeller in SAS-U is generally faster than the rapid predictive modeler in SODA - mainly owing to not fitting decision trees, logit models et c.