Hello there!
I was wondering if there’s a way to make “simple” decisions on a decision tree.
Let’s say I have this list adult.txt (3.8 MB)
It’s not the actual list, but it’s pretty similar to the actual one. I need to predict a numeric value, income, based on only two string types columns.
I need the decission tree to make simple decission that combines two previous columns. Example of a decission:
if Male AND married → X value
if Male and never-married … etc.
Technically, it is possible although the quality of the classifier will not be the best, as you already know
Here I suggest a trick that may solve your problem. You need to use the Tree Ensemble Learner because the Tree Decision Learner does not have an option to limit the levels of the Tree decision. Below you’ll find the configuration required for the Tree Ensemble Learner. It is configured to have a “single tree”, which makes it a “Decision Tree” instead of an Ensemble. The option “Limit Number of Levels (tree depth)”, set to value 2, simulates what you want to do.