any plans for supporting pred_contribs?
Currently XGBoost Python API support pred_contribs in Booster.predict. This option is useful for debugging some certain prediction cases because we can get contributions of each feature and find some explanation for the prediction result.
The API Doc http://xgboost.readthedocs.io/en/latest/python/python_api.html#xgboost.Booster.predict Related commit https://github.com/dmlc/xgboost/commit/6bd1869026eb66a542ccd8573f1e4b098a8cf7c6 Inspired by http://blog.datadive.net/interpreting-random-forests/
I'm wondering is there any plans for support this feature in this project? Thanks sincerely!
I've been working on this over the last few days. I have a non-functioning draft that I'm debugging. Let me know if you'd like to help.
I have a functioning contribution that I'm just about ready to push. I just need some help upgrading gradle as I have to run the contribution by the open-source overlords at my employer.
@bryan-woods Sorry for the late reply and thank you so much for the contribution! I'm not an expert about gradle, but I'm happy to give you as much help as I can. Let me know if you need my help.
@cocorosiekz I too am sorry for the late reply. I just turned my attention back to this topic. Here is the ticket I created: https://github.com/komiya-atsushi/xgboost-predictor-java/issues/30