Minimal Description Length Pruning

I would like more information about the implementation of Minimal Description Length Pruning (MDLP) in KNIME’s Decision Tree Learner. I have searched for details on how this is implemented and could not find any. There are conceptual articles on MDLP with quite complex mathematics, but no example of how the algorithm works.

I understand that if the implementation of MDLP in KNIME is proprietary, then no details can be provided. I would at least like to know if it is proprietary or to have some detail. I am planning to teach predictive analytics using KNIME and I am certain that one or more of my students will ask for an explanation.



Hi @acito,

great to read that you are planning to use KNIME in your lecture.

Are you in general familiar with the idea of MDL pruning? If not in my opinion slide 16 and 17 of this following slide deck give a nice explanation:

I’m not 100% sure how the cost for the tree encoding is calculated in KNIME, but the code is open source and you could have look at it:


This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.