When visualizing a regression tree in RapidMiner (that is, a decision tree with the least squares criterion to predict a numeric label), the branches are arranged such that the less-than branches are on the right and the greater-than branches are on the left. This is completely counterintuitive and impedes effective interpretation of the tree. Here is an illustration to make this point clearer:
Take the top branch, for example, the split on "jean". Everything where jean is > 0.126 is on the left, whereas everything where jean is ⩽ 0.126 is on the right. This is counterintuitive. I would expect that the left branch would be jean ⩽ 0.126 and the right branch would be jean > 0.126. This counterintuitive ordering is the case throughout the regression tree.
Is there a way to automatically flip the branch order by default? If not, I would like RapidMiner to reverse the order by default, since I think the reverse would make sense to most users. However, if there is a use case for the present default order, then perhaps there could be an option somewhere for the user to specify the preferred order.
Could anyone please help me with this, or how can I turn this into a feature request (or more like a minor bug fix, in my opinion).