Inferring feature importance with uncertainties with application to large genotype data

Estimating feature importance, which is the contribution of a prediction or several predictions due to a feature, is an essential aspect of explaining data-based models. Besides explaining the model itself, an equally relevant question is which features are important in the underlying data generatin...

Full description

Saved in:
Bibliographic Details
Published in:PLoS computational biology Vol. 19; no. 3; p. e1010963
Main Authors: Johnsen, Pål Vegard, Strümke, Inga, Langaas, Mette, DeWan, Andrew Thomas, Riemer-Sørensen, Signe
Format: Journal Article
Language:English
Published: United States Public Library of Science 01.03.2023
Public Library of Science (PLoS)
Subjects:
Artificial intelligence
Biobanks
Biology and Life Sciences
Computer and Information Sciences
Confidence intervals
Decomposition
Engineering and Technology
Estimation theory
Expected values
Game theory
Genotype
Genotype & phenotype
Genotypes
Genotyping Techniques
Machine learning
Neural networks
Obesity
Physical Sciences
Random variables
Resampling
Research and Analysis Methods
Synthetic data
Tree structures (Computers)
Uncertainty
Values
Norway
United Kingdom > UK
ISSN:1553-7358, 1553-734X, 1553-7358
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Estimating feature importance, which is the contribution of a prediction or several predictions due to a feature, is an essential aspect of explaining data-based models. Besides explaining the model itself, an equally relevant question is which features are important in the underlying data generating process. We present a Shapley-value-based framework for inferring the importance of individual features, including uncertainty in the estimator. We build upon the recently published model-agnostic feature importance score of SAGE (Shapley additive global importance) and introduce Sub-SAGE. For tree-based models, it has the advantage that it can be estimated without computationally expensive resampling. We argue that for all model types the uncertainties in our Sub-SAGE estimator can be estimated using bootstrapping and demonstrate the approach for tree ensemble methods. The framework is exemplified on synthetic data as well as large genotype data for predicting feature importance with respect to obesity.
Bibliography:new_version
ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
The authors declare that they have no competing interests.
ISSN:1553-7358
1553-734X
1553-7358
DOI:10.1371/journal.pcbi.1010963