Department of Mathematical Sciences
|DATE:||Thursday, Feb. 17, 2022|
|TIME:||1:15pm – 2:15pm|
|SPEAKER:||Shaofei Zhao, Binghamton University|
|TITLE:||Unbiased measurement of feature importance in tree-based methods|
In this article, the authors proposed a modification that corrects for split-improvement variable importance measures in Random Forest and other tree-based methods. These methods have been shown to be biased towards increasing the importance of features with more potential splits. The authors showed that by appropriately incorporating split-improvement as measured on out of sample data, this bias can be corrected yielding better screening tools.