[
https://issues.apache.org/jira/browse/SPARK-11730?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15022443#comment-15022443
]
Seth Hendrickson commented on SPARK-11730:
------------------------------------------
[~josephkb] Please see the paper below by Friedman, equations 44 and 45. He
proposes that variable importance should be the average across all trees for a
collection of trees.
https://statweb.stanford.edu/~jhf/ftp/trebst.pdf
Intuitively it would make sense to me to incorporate tree weights in the
feature importance, but I have found no instances either in theory or in
practice of this adjustment. Since R and scikit provide feature importance
according to the method above, I think it makes sense to stick to that
convention. Your thoughts are appreciated.
> Feature Importance for GBT
> --------------------------
>
> Key: SPARK-11730
> URL: https://issues.apache.org/jira/browse/SPARK-11730
> Project: Spark
> Issue Type: New Feature
> Components: ML, MLlib
> Reporter: Brian Webb
>
> Random Forests have feature importance, but GBT do not. It would be great if
> we can add feature importance to GBT as well. Perhaps the code in Random
> Forests can be refactored to apply to both types of ensembles.
> See https://issues.apache.org/jira/browse/SPARK-5133
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
