Gradient Boosted Trees: get full list of feature importances
Hi RapidMiner,
I'm using Gradient Boosted Trees (GBT) for a binary classifier with ~ 500 features, the result is good but I also want to see a full list of feature/variable importances instead of only 10 in the output model description (as in the screenshot attached). I noticed there's a similar question last year: http://community.rapidminer.com/t5/Product-Ideas/Gradient-Boosted-Trees-extract-feature-importance/idi-p/33066
Is it possible to do this in the current RM 7.5.001? Or do you have any idea of how to get this from the output GBT model? If you want my process and data to check, let me know and I will send it individually to you.
Thank you very much for your help!
P/S: I'm using a licenced version of RM Studio Large under my company account: chungsd@stee.stengg.com
Best,
phivu
GBT-output-model-description
Find more posts tagged with
Oh, I got it. Thank you very much Martin!
Hi,
Could you please explain what is the basis for ranking the importance of attributes in the GBT? For example, is it based on information gain or does it use a backward propagation/forward elimination approach such as the one SelectAttribute operator does? I would appreciate your answers. I would appreciate even more if you could provide me with an article or a webpage (hopefully from rapidminer documentation) that explains the mathematical logic for ranking the attribute importance in Gradient Boosted Trees (GBT)
Thanks,
@Pirehelokan, you should tag @mschmitz in your question, so that he can be alerted. Thanks Martin for your help. 
Altair EmployeeDear phivu,
This is being handled in 7.4 i think. The GBT has now a port called wei. This contains the very same list as a rapidminer weight object. That can be used to select attributes using Select by Weights or be transformed into an example set using weights to data.
Best,
Martin
Dear phivu,
This is being handled in 7.4 i think. The GBT has now a port called wei. This contains the very same list as a rapidminer weight object. That can be used to select attributes using Select by Weights or be transformed into an example set using weights to data.
Best,
Martin