Decision Tree - primary outcome

[snandiwada]

I am a novice learning SAS Enterprise Miner for Predictive Modeling. I am trying to learn about Decision Tree model and was wondering how you would change the primary outcome that the Tree starts with. I am looking at the PVA97NK data and when I try to run the interactive model for a binary target (0 and 1), the tree is being generated for target =0. How can I change the primary outcome to be 1 instead of 0.

[Cynthia_sas]

Hi:
Are you working in the Applied Analytics using SAS Enterprise Miner course? If so, what Lesson and practice are you doing with the PVA97NK data?
Thanks, in advance, Cynthia

[snandiwada]

I am in Lesson 4: decision trees

Looking at Constructing Decision Trees Interactively

TargetB is the target which is binary  and  the lesson demo shows that split is based on target = 1 but the decision tree in my project diagram shows that outcome = 0 is what is being predicted. 

[JThompson · Accepted solution]

Hi.  I'm one of the instructors for that class.  Either the issue is simply an "interpretation" one or I need more information about your question.  I'm thinking it may just be a misinterpretation of what the software is showing.  Enterprise Miner will by default always model the 1's given a binary target. In fact, there is not a property to change the target level being predicted.  It will always model the 1's. If it appears that certain nodes in the tree are predicting 0's, that is likely the decision prediction coming from that node.  Meaning, the subset of the data that the node in question is describing has found a subset of the data where the majority of cases are 0's, so any decision prediction based on that node is 0.  Even though in the training data, some 1's may have ended up in that node, the final prediction is 0, since that is the majority for the cases that end of there. Some nodes will predict the outcome as 1 and some nodes predict the outcome as 0.  I think you may just be looking at one of the nodes where the prediction is 0.

If this does not answer your question, then I may need more info on what you are seeing.  In this case, please provide a screen capture or a detailed description of exactly what you are seeing in the node while doing the exercise and how it is different from what you see in the notes.