Creating Unbiased Machine Learning Models by Design

Unintended bias against protected groups has become a key obstacle to the widespread adoption of machine learning methods. This work presents a modeling procedure that carefully builds models around protected class information in order to make sure that the final machine learning model is independen...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Joseph L. Breeden, Eugenia Leonova
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	unintended bias fair lending multihorizon survival models machine learning Risk in industry. Risk management HD61 Finance HG1-9999
Acceso en línea:	https://doaj.org/article/b9bac4d1e3e045759642b5816772143f
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:b9bac4d1e3e045759642b5816772143f
record_format	dspace
spelling	oai:doaj.org-article:b9bac4d1e3e045759642b5816772143f2021-11-25T18:08:50ZCreating Unbiased Machine Learning Models by Design10.3390/jrfm141105651911-80741911-8066https://doaj.org/article/b9bac4d1e3e045759642b5816772143f2021-11-01T00:00:00Zhttps://www.mdpi.com/1911-8074/14/11/565https://doaj.org/toc/1911-8066https://doaj.org/toc/1911-8074Unintended bias against protected groups has become a key obstacle to the widespread adoption of machine learning methods. This work presents a modeling procedure that carefully builds models around protected class information in order to make sure that the final machine learning model is independent of protected class status, even in a nonlinear sense. This procedure works for any machine learning method. The procedure was tested on subprime credit card data combined with demographic data by zip code from the US Census. The census data serves as an imperfect proxy for borrower demographics but serves to illustrate the procedure.Joseph L. BreedenEugenia LeonovaMDPI AGarticleunintended biasfair lendingmultihorizon survival modelsmachine learningRisk in industry. Risk managementHD61FinanceHG1-9999ENJournal of Risk and Financial Management, Vol 14, Iss 565, p 565 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	unintended bias fair lending multihorizon survival models machine learning Risk in industry. Risk management HD61 Finance HG1-9999
spellingShingle	unintended bias fair lending multihorizon survival models machine learning Risk in industry. Risk management HD61 Finance HG1-9999 Joseph L. Breeden Eugenia Leonova Creating Unbiased Machine Learning Models by Design
description	Unintended bias against protected groups has become a key obstacle to the widespread adoption of machine learning methods. This work presents a modeling procedure that carefully builds models around protected class information in order to make sure that the final machine learning model is independent of protected class status, even in a nonlinear sense. This procedure works for any machine learning method. The procedure was tested on subprime credit card data combined with demographic data by zip code from the US Census. The census data serves as an imperfect proxy for borrower demographics but serves to illustrate the procedure.
format	article
author	Joseph L. Breeden Eugenia Leonova
author_facet	Joseph L. Breeden Eugenia Leonova
author_sort	Joseph L. Breeden
title	Creating Unbiased Machine Learning Models by Design
title_short	Creating Unbiased Machine Learning Models by Design
title_full	Creating Unbiased Machine Learning Models by Design
title_fullStr	Creating Unbiased Machine Learning Models by Design
title_full_unstemmed	Creating Unbiased Machine Learning Models by Design
title_sort	creating unbiased machine learning models by design
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/b9bac4d1e3e045759642b5816772143f
work_keys_str_mv	AT josephlbreeden creatingunbiasedmachinelearningmodelsbydesign AT eugenialeonova creatingunbiasedmachinelearningmodelsbydesign
_version_	1718411580612280320

Creating Unbiased Machine Learning Models by Design

Ejemplares similares