CoChat

Feature Importance in Gradient Boosting Trees with Cross-Validation Feature Selection

Open access

The DOI matches its Crossref record; 37 of the 39 references it cites were found in a registry or an index.

Abstract

Gradient Boosting Machines (GBM) are among the go-to algorithms on tabular data, which produce state-of-the-art results in many prediction tasks. Despite its popularity, the GBM framework suffers from a fundamental flaw in its base learners. Specifically, most implementations utilize decision trees that are typically biased towards categorical variables with large cardinalities. The effect of this bias was extensively studied over the years, mostly in terms of predictive performance. In this work, we extend the scope and study the effect of biased base learners on GBM feature importance (FI) measures. We demonstrate that although these implementation demonstrate highly competitive predictive performance, they still, surprisingly, suffer from bias in FI. By utilizing cross-validated (CV) unbiased base learners, we fix this flaw at a relatively low computational cost. We demonstrate the suggested framework in a variety of synthetic and real-world setups, showing a significant improvement in all GBM FI measures while maintaining relatively the same level of prediction accuracy.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/e24050687

Cites

37 of 37 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 3 Scaled FI on the Amazon data-set, across 30 folds. Variables are ordered by their cardinality, from left to right, in descending order.
Figure 3 Scaled FI on the Amazon data-set, across 30 folds. Variables are ordered by their cardinality, from left to right, in descending order.CC BY · PubMed Central
Figure 5 Scaled FI on the Criteo CTR data-set, across 30 folds. Variables are ordered by their cardinality, from left to right, in descending order. For visualization, FI results for features with low cardinality that have small FI values are grouped.
Figure 5 Scaled FI on the Criteo CTR data-set, across 30 folds. Variables are ordered by their cardinality, from left to right, in descending order. For visualization, FI results for features with low cardinality that have small FI values are grouped.CC BY · PubMed Central
Figure A1 SHAP FI for the null case.
Figure A1 SHAP FI for the null case.CC BY · PubMed Central
Figure A2 SHAP FI for the power case.
Figure A2 SHAP FI for the power case.CC BY · PubMed Central
Figure A3 FI on the Amazon and Criteo CTR data-sets, including SHAP FI.
Figure A3 FI on the Amazon and Criteo CTR data-sets, including SHAP FI.CC BY · PubMed Central
Figure 1 Scaled Gain FI (left) and PFI (right) for the null case experiment where all displayed features are uninformative.
Figure 1 Scaled Gain FI (left) and PFI (right) for the null case experiment where all displayed features are uninformative.CC BY · PubMed Central
Figure 2 Scaled Gain FI (left) and PFI (right) for the power case experiment where only X1 is informative. Since X1 is informative, it is on a different scale and it is therefore omitted to improve visualization. Its mean values are 0.186, 0.181, 0.166 and 0.145 for Vanilla GBM, LGBM, CatBoost and CVB, respectively.
Figure 2 Scaled Gain FI (left) and PFI (right) for the power case experiment where only X1 is informative. Since X1 is informative, it is on a different scale and it is therefore omitted to improve visualization. Its mean values are 0.186, 0.181, 0.166 and 0.145 for Vanilla GBM, LGBM, CatBoost and CVB, respectively.CC BY · PubMed Central
Figure 4 Error (log-loss) on the Amazon data-set, across 30 folds, with and without the Resource variable.
Figure 4 Error (log-loss) on the Amazon data-set, across 30 folds, with and without the Resource variable.CC BY · PubMed Central
Figure 6 Error (log-loss) on the Criteo data-set, across 30 folds, with and without the following variables: Device_ip, Device_id, Device_model.
Figure 6 Error (log-loss) on the Criteo data-set, across 30 folds, with and without the following variables: Device_ip, Device_id, Device_model.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 37 references

Research works printed but not linked: 1

  1. Loh W.Y. Regression tress with unbiased variable selection and interaction detection Stat. Sin. 2002 12 361 386

Cited by

See all 186

Metadata

Authors
Adler, A. I., & Painsky, A.
Affiliations
Tel Aviv University (IL)
Corresponding author
Painsky, Amichai
Venue
Entropy
Publisher
MDPI AG
Published
2022-05-13
Volume
24
Issue
5
Pages
687
ISSN
1099-4300 (ISSN-L)
OpenAlex
W3200824799
Cited by
149
References
39
Citations by year
2026: 35 · 2025: 59 · 2024: 32 · 2023: 18 · 2022: 5
FWCI
9.38
Type
journal article
Language
en
Open access
Locations
  • Entropy (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University) (cc-by)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
  • Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Inference
Keywords
gradient boosting, feature importance, categorical variables, decision tree, tabular data

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.