🎉 🎉 RAPIDMINER 9.10 IS OUT!!! 🎉🎉

Download the latest version helping analytics teams accelerate time-to-value for streaming and IIOT use cases.

CLICK HERE TO DOWNLOAD

Performance of individual Gradient Boosted Trees

Muhammed_Fatih_Muhammed_Fatih_ Member Posts: 93  Maven
Dear Community, 

I have executed the GBT tutorial process and could see that an overall Performance Vector was generated. I assume that the underlying performance values are representative for the sum of the ten GBTs which were generated, right? 

Is there a possbility to see the performance development of the intermediate steps or does the algorithm "only" calculates an overall score? So probably, what is the performance of Tree 5? 



Best regards

Fatih  

Answers

  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,158  RM Data Scientist
    Hi Fatih,
    i think there is a bit of a misconception. Those trees are not parallel to one another but depend on the proceeding ones. If you want to use a subset of the trees, you would need to evaluate the first 1-k trees compared to the first 1-k+m trees. I don't see a reason to evaluate trees individually.

    ABoth things are afaik not easily doable.

    Best,
    Martin
    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
  • Muhammed_Fatih_Muhammed_Fatih_ Member Posts: 93  Maven
    Hi @mschmitz

    thank you for your answer!

    Does this mean that the Performance Vector calculates the values for the last (in this case:10th) Decision Tree which is sequentially optimised?  

    Best regards! 

    Fatih
  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,158  RM Data Scientist
    Hi,

    maybe have a look at this kb article: https://community.rapidminer.com/discussion/36379/a-practical-guide-to-gradient-boosted-trees-part-i-regression this should help. Happy to help for further questions.

    Best,
    Martin

    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
Sign In or Register to comment.