Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: From unbiased MDI Feature Importance to Explainable AI for Trees
(Submitted on 26 Mar 2020 (v1), last revised 30 Sep 2021 (this version, v4))
Abstract: We attempt to give a unifying view of the various recent attempts to (i) improve the interpretability of tree-based models and (ii) debias the the default variable-importance measure in random Forests, Gini importance. In particular, we demonstrate a common thread among the out-of-bag based bias correction methods and their connection to local explanation for trees. In addition, we point out a bias caused by the inclusion of inbag data in the newly developed explainable AI for trees algorithms.
Submission history
From: Markus Loecher [view email][v1] Thu, 26 Mar 2020 17:16:58 GMT (114kb,D)
[v2] Wed, 15 Apr 2020 17:45:50 GMT (122kb,D)
[v3] Sun, 24 May 2020 15:44:32 GMT (127kb,D)
[v4] Thu, 30 Sep 2021 14:35:53 GMT (126kb,D)
Link back to: arXiv, form interface, contact.