Xgboost: Meaning of feature importance score

Created on 30 May 2015  ·  4Comments  ·  Source: dmlc/xgboost

Do they have an interpretable semantics? How are they calculated? Does higher mean better?

To clarify, I'm using cls.booster().get_fscore() to get the scores.

picture FabHan
👍2

Most helpful comment

that means these feature never get selected into the trees

picture tqchen on 2 Jun 2015
👍2

All 4 comments

Also, get_fscore() returns fewer features than the number of features in the training data. I have 98 features and get_fscores() return scores of 71 features.

FabHan picture FabHan on 30 May 2015

The higher the better, get_fscore returns number of occurance of features in the ensemble

tqchen picture tqchen on 2 Jun 2015
👍2

Does it use their levels in the tree as weights?

Also, do you have an explanation for the situation in my second question?

Thanks.

FabHan picture FabHan on 2 Jun 2015

that means these feature never get selected into the trees

tqchen picture tqchen on 2 Jun 2015
👍2
Was this page helpful?
0 / 5 - 0 ratings

Related issues

choushishi picture choushishi  ·  3Comments
tqchen picture tqchen  ·  4Comments
vkuznet picture vkuznet  ·  3Comments
yananchen1989 picture yananchen1989  ·  3Comments
trivialfis picture trivialfis  ·  3Comments