{
  "id": "1903.05179",
  "version": "v1",
  "published": "2019-03-12T19:52:53.000Z",
  "updated": "2019-03-12T19:52:53.000Z",
  "title": "Unbiased Measurement of Feature Importance in Tree-Based Methods",
  "authors": [
    "Zhengze Zhou",
    "Giles Hooker"
  ],
  "categories": [
    "stat.ML",
    "cs.LG"
  ],
  "abstract": "We propose a modification that corrects for split-improvement variable importance measures in Random Forests and other tree-based methods. These methods have been shown to be biased towards increasing the importance of features with more potential splits. We show that by appropriately incorporating split-improvement as measured on out of sample data, this bias can be corrected yielding better summaries and screening tools.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-03-12T19:52:53.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "tree-based methods",
      "feature importance",
      "unbiased measurement",
      "split-improvement variable importance measures",
      "random forests"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}